Home

Ali-Master-Thesis

1. 004 75 A 2 1 Compilation Tips for Windows Users oa aa a 76 Bibliography 17 Glossary 79 List of ACFONYMS 2 2 is 6020 ia ee ee ey OES a a eS 79 List of Symbols is A oe Se A ee E n 80 Ali Osman ISIK Master of Science Thesis Acknowledgements First I would like to thank my supervisor Prof dr R Babuska for giving me the opportunity to do my MSc Thesis in DCSC and for his feedbacks on my final report Special thanks to my daily supervisor Dr G A D Lopes for his tremendous contribution to the Distributed Robotics Library and for motivating me during the thesis I also want to thank to A Si monetto for giving significant feedbacks and providing the testbed that I used to test the Distributed Robotics Library Last but not the least I would like to thank my family and friends for their support Delft University of Technology Ali Osman ISIK November 25 2010 Master of Science Thesis Ali Osman ISIK vi Acknowledgements Ali Osman ISIK Master of Science Thesis Chapter 1 Introduction Distributed Control Systems DCSs are a class of systems designed to control distributed processes or complete manufacturing systems by using decentralized elements or subsystems In centralized control a core controller is responsible for the pre planning task assignment and supervision of the task coordination However in DCSs agents are capable of sensing acting communicating and they contribute to the task solution all togeth
2. 5 In order to create documentation files run make doc command 6 In thesis_library documentation folder you can find html and latex documentation files of the library A 2 1 Compilation Tips for Windows Users If you are using a Windows system it is strongly recommended to install cygwin cross compiler in order to use CMake and Doxygen properly Then it is possible to compile the library by following the same steps explained above Note that the executable files are windows executable now and can be run on Windows systems by either adding location of the cygwinl dll file to the system path or adding the file to the same folder with the executable Ali Osman ISIK Master of Science Thesis Bibliography 1 H Ben tez P rez and F Garc a Nocetti Reconfigurable Distributed Control Secaucus NJ USA Springer Verlag New York Inc 2005 2 J Galdun L Takac J Ligus J Thiriet and J Sarnovsky Distributed control systems reliability Consideration of multi agent behavior in 6th International Symposium on Applied Machine Intelligence and Informatics 2008 SAMI 2008 pp 157 162 2008 3 J Aguilar M Cerrada G Mousalli F Rivas and F Hidrobo A multiagent model for intelligent distributed control systems in Knowledge Based Intelligent Information and Engineering Systems R Khosla R J Howlett and L C Jain eds vol 3681 of Lecture Notes in Computer Science pp 1
3. By passing this pointer to the interface function in listoffunctions h we enable this interface function to invoke the real function which will do the job The second and third message handlers are robots own member functions thus we passed a NULL pointer in the entity input After preparing the library we are ready to test functions of the library Note that we run the same application in all PCs but the following functions will only be called by ALPHA1 After running the applications on all PCs they took care of advertisements automatically and all the PCs are ready to set up a connection between each other First function invoked by ALPHA1 is GetRobotCommList function to see the available robots and connection information in the network Here is the output of this function call Robot_2 ALPHA1 console New Advertisement detected from ALPHA3 New Advertisement detected from ALPHA4 New Advertisement detected from ALPHA5 New Advertisement detected from ALPHA2 List of the robots in the network ALPHA3 Not Connected ALPHA4 Not Connected ALPHA5 Not Connected ALPHA2 Not Connected As we see from the output after running the application ALPHA1 detected advertisements from other robots and we see all the robots in the network none of them are connected to ALPHA1 Then ALPHA1 sends connection requests to ALPHA2 and ALPHA5 by using SendConnectionRequest function and it prompts the list of robots again The output is as
4. It allows the user the transmission whenever there is data to be sent With a LAN a user can listen the channel while transmitting and can understand if there is a collision by a hardware feedback system If listening while transmitting is not possible i e WLAN a user can detect the collision by some acknowledgments If the frame of a user is destroyed the sender waits random amount of time and tries to transmit it again These kind of systems are called contention systems since there is a contention between multiple users In this kind of a system collision is of course inevitable If we look at the efficiency of the ALOHA network the best channel utilization we can get is 18 percent 12 Since this result is not very encouraging a better method is presented as slotted ALOHA The only difference between ALOHA and slotted ALOHA is that time is divided into discrete intervals of one frame length In this method a transmission is only allowed at the beginning Master of Science Thesis Ali Osman ISIK 12 Communication in Networks of the time slots In ALOHA network the vulnerable period was two frame size however in slotted ALOHA it is one frame size So the efficiency is doubled and best channel utilization is 37 percent This is still not very encouraging result In Local Area Networks LANs it is possible for users to detect what others are doing So to improve the performance Carrier Sense Multiple Access CSMA protocol is presented
5. This layer is responsible for formatting of and the delivery of information to the Application Layer for further process We can list some of the services of the Presentation Layer as follows e Providing common representation of data to be transferred between application entities That relieves application entities form concerning any problem of common representa tion of information and gives application entities syntax freedom e Encryption can also be done in this layer e It is responsible for selecting transfer syntax between application entities In many applications there is no distinction between the Application Layer and the Pre sentation Layer Most of the Application Layer protocols have Presentation Layer aspects too American Standard Code for Information Interchange ASCII is the most prominent Presentation Layer standard Layer 7 Application Layer The Application Layer is the last layer in OSI model and closest to the end user This is the OSI layer that interacts with the software applications implementing communication components The services provided by this layer can be listed as follows e Identifying the communication partners and determining the data to transmit e Determining whether sufficient network resources available for the communication e Deciding whether the requested communication exists e It synchronizes the communication In this layer there are a lot of protocols The most prominent ones are H
6. an active scan message to see the devices in its neighborhood or it can listen beacons from other devices After scanning period a device can send an association request to the PAN it wants to join and it gets an address from the corresponding coordinator and joins the network Deassociation can be done either by decision of a coordinator or device itself In either case device and coordinator removes all references of each other When a device associates with a PAN it has to synchronize with the coordinator In beacon enabled systems Ali Osman ISIK Master of Science Thesis 2 3 High level Communication Protocols 31 synchronization is performed by receiving and decoding beacon frames In the other systems synchronization is performed by polling the coordinator for data Synchronization is updated in every transmission When timeout occurs device is considered as an orphan and orphan notification commands are sent for re synchronization Frame Structures There are four types of frames beacon data acknowledgment control The general MAC frame structure is shown in Figure 2 8 The descriptions of all types of frames and all the fields are beyond the scope bytes 2 size Frame A Frame control number PAN ID ri PAN ID pd Payload Address fields MHR MAC MFR payload Figure 2 8 General Frame Structure in ZigBee As we stated these were the layers standard by IEEE 802 15 4 The key definition in ZigBee is called the ZigBee Devi
7. e Providing quality of service parameters such as error rate service availability transmis sion rate and transmission delay For different medium access protocols different physical layer specifications are presented In these specifications type of the wire transceivers transceiver cables connectors hub types repeater types terminators distance limitations are varied The most prominent physical layer protocols are RS232 USB Ethernet Physical Layer Token Ring Physical Layer etc These protocols are important but out of scope of this thesis In Wireless Local Area Network WLAN IEEE 802 11 five different transmission techniques can be used Infrared Frequency Hopping Spread Spectrum FHSS Direct Sequence Spread Spectrum DSSS Orthogonal Frequency Division Multiplexing OFDM and High Rate Direct Sequence Spread Spectrum HR DSSS are used for physical data transmission Ali Osman ISIK Master of Science Thesis 2 2 Modeling of Communication Libraries 11 Layer 2 Data Link Layer The Data Link Layer DLL is the second level in OSI model The DLL provides functional and procedural means for connectionless mode among network entities and for connection mode for the establishment maintenance and release data link connections among network entities and for the transfer of data link service data units A data link connection is built upon one or several physical connections 11 The DLL can be divided into two sublayers
8. gt ROE Cee A IS a Az slave sy a j O slave y Y i Q A i f O 6 i 1 1 H H n i master j 1 O i f o slave f master 6 Ye e slave slave Soe sf aS Oo slave PA Re eee Figure 2 6 A Bluetooth Scatternet Application Applications Profiles layer Middleware er nish Ss ca Paty a ye Logical link control adaptation protocol Data link layer Physical layer Baseband Figure 2 7 The 802 15 version of the Bluetooth protocol architecture Master of Science Thesis Ali Osman ISIK 26 Communication in Networks e Class 1 has a range around 100 meters and uses 100 mW power It can be used for Bluetooth Access Points e Class 2 has a range around 10 meters and uses 2 5 mW power It can be used for PCs and portable devices e Class 3 has a range around 1 meter and uses 1 mW power It can be used for low power devices Bluetooth Baseband Layer This layer is the most similar layer to the MAC layer in Bluetooth Physical Characteristics The aim of this layer is to map logical channels to physical ones As we stated earlier the logical channel is selected among 79 available RF frequencies in the 2 4 GHz band and pseudo random hopping sequence is used to select the channels The nodes that use the same sequence form a piconet Since 1600 hops sec is used time is divided into slots of length 625 psec each by using the clock of the master Master and s
9. AP sends a random number station encrypts random number and sends back to AP After AP decrypts the number and matches the key association is completed e WEP Integrity When a frame is sent to a station station appends 32 bit Integrity Check Vector ICV to the frame e WEP Confidentiality Since the same shared key is used in every transmission an addi tional 24 bit Initialization Vector IV is appended to each frame Since IV is plaintext anyone sniffing the WLAN can see it Thus it should not be used twice However with 24 bits 16 777 216 possible numbers can be generated and in 5000 transmissions possibility of generating the same number is 0 50 So anyone can listen and capture the data if the same key is generated Since WEP supplies very unsecured security services latter versions of Wi Fi uses WPA Wi Fi Protected Access or WPA2 In these standards 256 bit key is used and since these keys are vulnerable to cracking further protection is applied Up to now we have seen three different wireless technologies Wi Fi Bluetooth and ZigBee In Figure 2 9 we can see the comparison of these wireless technologies 2 3 4 Zero Configuration Networking Communication across the planet is supplied by TCP IP without worrying about the type of physical connection such as Ethernet wireless etc This is the inspiration of solving the local communication problem For local purposes TCP IP is the way too powerful and complicated techniq
10. In this protocol channels listen for carriers and act accordingly One type is 1 persistent CSMA In this protocol when a station has something to send it senses the channel If channel is busy it waits the channel to become idle When it is idle it transmits the frame If a collision is detected the station waits random amount of time and retransmits the frame It gives a better channel utilization than slotted ALOHA However like ALOHA and slotted ALOHA under heavy traffic this method is also very inefficient due to high probability of collisions A better solution under heavy traffic is non persistent CSMA In this protocol if a station finds the channel idle it transmits If the channel is busy it waits random amount of time and senses the channel again Consequently it gives a better channel utilization under heavy traffic but longer delays Both persistent and non persistent protocols can detect the collision but they still finish their transmissions To improve these protocols a better protocol is presented called Carrier Sense Multiple Access With Collision Detection CSMA CD CSMA with collision detection The improvement of this protocol is whenever a channel senses a collision it directly terminates the transmission waits a random period of time and tries again In Ethernet IEEE 802 3 this method is used In this protocol the key issue is the time required to detect a collision in the worst case scenario Note that the colli
11. OSI model are concerned with application issues and the lower levels are concerned with transmission issues 10 We can list the services provided by the Transport Layer as follows e Segmentation of upper level services and reassembly of lower level packets By segmen tation it encapsulates application data blocks to data segments suitable for transmission on the Network Layer e When the Network Layer does not guarantee the delivery of packets in the correct order Transport Layer provides it by giving numbers to each packet e It guarantees the reliable transmission of the data from origin to destination It estab lishes a high level of error checking and it recovers errors by retransmission of erroneous packets e Maintenance of information flow control when it is not provided by the Network Layer The Transport Layer protocols can be classified into five groups e Class 0 TPO This is the simplest protocol for connection oriented networks It performs no error checking or flow control Master of Science Thesis Ali Osman ISIK 18 Communication in Networks Class 1 TP1 It can monitor for transmission errors and recover the error by resending it So it provides reliable service Class 2 TP2 It can monitor for transmission errors and provides flow control between Transport Layer and Session Layer Class 3 TP3 In addition to the functions of TP1 and TP2 it has more reliable recovery mechanism Class 4 TP4 In
12. SENDER gt ALPHA3 lt SENDER gt lt SENDER gt ALPHA1 lt SENDER gt lt FUNCTION gt PrintMessage lt FUNCTION gt lt HANDLER gt lt GetTime gt lt FUNCTION gt lt SendToOtherRobot gt Message sent to ALPHA1 lt PrintMessage gt lt STRING gt Localtime is 15 14 29 lt STRING gt lt PrintMessage gt Output of ALPHA Message sent to ALPHA4 lt SendToOtherRobot gt lt RECEIVER gt ALPHA5 lt RECEIVER gt lt FUNCTION gt lt GetTime gt lt HANDLER gt lt SENDER gt ALPHA3 lt SENDER gt lt SENDER gt ALPHA1 lt SENDER gt lt FUNCTION gt PrintMessage lt FUNCTION gt lt HANDLER gt lt GetTime gt lt FUNCTION gt lt SendToOtherRobot gt Message sent to ALPHA1 lt PrintMessage gt lt STRING gt Localtime is 15 14 29 lt STRING gt lt PrintMessage gt Output of ALPHA5 Message sent to ALPHA4 lt SendToOtherRobot gt lt RECEIVER gt ALPHA3 lt RECEIVER gt lt FUNCTION gt lt SendToOtherRobot gt lt RECEIVER gt ALPHA1 lt RECEIVER gt lt FUNCTION gt lt PrintMessage gt Master of Science Thesis Ali Osman ISIK 70 Experimental Evaluation of Distributed Robotics Library lt STRING gt Localtime is 15 14 29 lt STRING gt lt PrintMessage gt lt FUNCTION gt lt SendToOtherRobot gt lt FUNCTION gt lt SendToOtherRobot gt In outputs of ALPHA and ALPHA4Z we see that the received XML messages are relayed to the receiver In the output of ALPHAJ the received message calls the Get
13. The upper sublayer is the Logical Link Control LLC layer and the lower sublayer is the Media Access Control MAC layer Media Access Control Layer The MAC sublayer acts as an interface between the physical layer and the LLC This layer is not required for a full duplex point to point communication but for the multi agent networks We can list the services of this sublayer as follows e Unique addressing of each device in the network This procedure is called MAC ad dressing In Ethernet addressing 48 bits of addressing is used e It provides multiple access protocols for channel access control These protocols make it possible for several nodes to share the same medium The most prominent protocols used for medium access will be explained in the following subsections e It is responsible for frame synchronization such as time based character counting byte stuffing and bit stuffing e It is responsible for switching queuing or scheduling of data packets in the local area network e It controls quality of service by controlling the parameters provided by physical layer e When there is no physical network it establishes a Virtual Local Area Network VLAN Now we will look into two of most prominent multiple access control protocols in detail Carrier Sense Multiple Access with Collision Detection CSMA CD The first protocol solving the channel allocation problem was the ALOHA system The idea behind this protocol is very simple
14. a notebook to connect to a phone without wires Fax Profile allows wireless fax machines to send and receive data from phones Cordless Telephony Profile connects the handset of a cordless telephone to the base station Intercom Profile allows two telephones to be used as walky talkies Headset Profile allows hands free communication with a headset Object Push Profile allows to exchange simple objects between two bluetooth devices File Transfer Profile is a more general file transfer facility e Synchronization Profile provides synchronization of a computer and a PDA 2 3 2 ZigBee One can categorize WPANSs as low data rate medium data rate and high data rate Bluetooth was an example of medium data rate WPANs To supply low data rate specifications IEEE 802 15 4 is designed and it specifies physical and MAC layer protocols ZigBee is the archi tecture developed on top of the IEEE 802 15 4 reference stack model and takes full advantage of its powerful physical radio layer 14 Thus Zigbee provides security discovery profiling kind of services on top of IEEE 802 15 4 standard ZigBee is specified to address low cost low power lightweight wireless network solutions By providing long battery life for devices in the networks it is especially used in wireless sensor networks The network layer of ZigBee supports three different topologies star tree and generic mesh networks The commonly used one is mesh topology which enables
15. addition to the functions of TP3 it can be used for connectionless networks and it has more extensive error monitoring and recovery It is the type closest to the TCP protocol In the following subsections we will explain two protocols used by Internet User Datagram Protocol UDP and User Datagram Protocol UDP User Datagram Protocol UDP is a connectionless protocol which provides a way for ap plications to send encapsulated IP datagrams and send them without a connection 12 UDP does not provide guarantee for message delivery so it is also called Unreliable Datagram Proto col However it provides application multiplexing via port numbers and integrity verification via checksum UDP transmits segments consisting of 8 byte header followed by the payload UDP header consists 4 fields e The source port is needed when a reply must be sent back to the source When it is not used it is assumed to be 0 e The destination port is required and identifies the destination port e 16 bit length field identifies the length of the datagram in bytes e Checksum is optional and it is not used in IPv4 while it is used in IPv6 UDP is a very simple protocol and it does not do flow control error control or retransmission of a bad segment All it does is providing an interface to IP protocol with demultiplexing processes using ports If an application requires reliability it should use Transmission Control Protocol TCP protocol UDP is especiall
16. as we see in Figure 2 6 14 Master node is responsible for chan nel allocation communication establishment setting the piconet synchronization clock and deciding for the frequency hopping sequence FHS Slaves cannot talk directly each other except discovery phase A slave can be part of different piconets and a master can be slave of another piconet as well In Bluetooth communication layering structure is slightly different than OSI model EEE 802 15 version of Bluetooth protocol architecture is given in Figure 2 7 12 Now we look into these layers in detail Bluetooth Radio Layer Since this is a physical layer it moves bits between master to slave The radio technology used is called frequency hopping spread spectrum with 1600 hops sec In this method channels are defined in spread spectrum and every transmission is done in different frequency channel Bluetooth operates at 2 4 GHz ISM band and except France and Spain this band is divided into 79 channels of 1 MHz each In Spain and France it is divided into 23 channels of 1 Mhz each Until Version 2 0 Gaussian frequency shift keying GFSK is used for frequency modulation and with 1 bit per hertz 1 Mbps data rate can be reached In 2 0 and later versions different methods are used and 2 or 3 Mbps data rates are reached There are 3 different types of devices in Bluetooth Ali Osman ISIK Master of Science Thesis 2 3 High level Communication Protocols 25 bridge A A Pee
17. communication network and exchanging internal data between netbooks Chapter 5 gives the overview of the library and discusses some possible future work Master of Science Thesis Ali Osman ISIK 4 Introduction Ali Osman ISIK Master of Science Thesis Chapter 2 Communication in Networks The purpose of this chapter is to give a reference guide to networking Most of the sections in this chapter serve as background information for the remainder of the thesis and can be safely skipped for a less computer science inclined reader However they are important in presenting a general idea about networking and giving a better understanding about the design of the Distributed Robotics Library DRL that we developed To maintain the fidelity to the original referenced document some texts are cited verbatim Since we are dealing with robotic networks in this thesis our focus is on Multi Agent Systems MASs and their communication properties In Section 2 1 we start by giving information about some properties of physical agents Then we explain how communication is modeled between agents with graph theoretic approach In Section 2 2 we explain the 7 layer Open Systems Interconnection OSI model which is the standard model for communication systems In this section we also give example protocols to the each layer In Section 2 3 we explain Bluetooth ZigBee Wi Fi which is used in this version of the library and Zero Configuration Networkin
18. into XML format which can be seen in the previous chapter After creation of a UDP socket the encoded data is sent to the IP address 255 255 255 255 and port number ADV_PORT over this UDP socket IP address 255 255 255 255 is used to send the data to the all Robots on the same subnet This process is repeated in each second Master of Science Thesis Ali Osman ISIK 50 Implementation of the Distributed Robotics Library ListenToAdvertisements This is a thread which starts running with the creation of the DistributedRoboticsLib it creates a UDP socket to listen ADV_ PORT and it listens this socket forever Once it receives a message it decodes the message and checks if the advertising Robot is already in the robot_comm_ list If yes first it first checks for an IP change If IP address of the Robot has changed RobotComm is removed and added with the new IP If RobotComm already exists with the same IP it sets the value of watch_dog variable to 3 in order to keep RobotComm alive If not it creates a new RobotComm and adds to the list Then it goes back to advertisement listening mode ListenConnectionRequests This is a thread which starts running with the creation of the Robot it creates a TCP server socket to listen ADV_PORT and it waits for the connection to the server forever Once it detects a connection attempt to the server it waits for the short information message In this version of the library advertisement message
19. is used for the short information message Once it receives the information message of the robot which is trying to connect it decodes the message and checks if the corresponding RobotComm is already in the robot_comm_ list If not it creates a new RobotComm and adds to the list Then it generates a random port number to communicate with the robot and sends this port number to the Robot which is trying to connect Then it calls StartCommunicationAsServer method of RobotComm with the new port number Then it goes back to connection waiting mode SendConnectionRequest This method sends connection request to a particular hostname in the network First it tries to find corresponding RobotComm If there is no RobotComm with this hostname connection request is ignored If RobotComm is found port number of the Robot is retrieved and short information is sent to the Robot In case the target robot is not listening the connection requests for the moment it keeps on sending the connection request until the request is sent Then the method starts listening the new port number which will be created by the server Robot Once it receives the new port number it calls StartCommunicationAsClient method of corresponding RobotComm with the new port num ber StopCommunication This robot is used to stop communication with a Robot First it finds the corresponding RobotComm Then it calls StopCommunication method of RobotComm The rest will be handled by Rob
20. of ALPHA2 A1i ALPHA2 console New Advertisement detected from ALPHA4 New Advertisement detected from ALPHA1 New Advertisement detected from ALPHA3 New Advertisement detected from ALPHA5 Connection Request detected from ALPHA1 Server started Message received from ALPHA1 lt MoveFirstLeg gt lt X gt 15 lt X gt lt Y gt 20 lt Y gt Master of Science Thesis Ali Osman ISIK 60 Experimental Evaluation of Distributed Robotics Library lt MoveFirstLeg gt First Leg is moved to X 15 Y 20 Likewise ALPHA 1 it also detected advertisements from the other robots Then it detected a connection request which was sent by ALPHA1 and it started server client communica tion process as a server Then it receives the message and calls the function with desired parameters As we see on the last line first leg of VirtualRobot is moved to desired position In figure 4 1 we explain the communication process better with the steps This figure also clarifies why the VirtualRobot pointer was passed while creating message handler in the first paragraph of this section List of ALPHA2 s Functions MoveFirstLeg GetTime 4 MoveFirstLeg of VirtualRobot is called ALPHA2 PrintMessage VirtualRobot 1 Send XML message to ALPHA2 3 MoveFirstLeg is called Robot2Comm 5 First Leg of VirtualRobot is moved Client 2 Sending XML message Robot1Comm Server Figure 4 1 XML message sending proce
21. protocol as Application Layer which includes the OSI Application Layer Presentation Layer and the Session Layer Application Layer of Advertisement Process Each Robot using the DRL is able to send and receive advertisements by using multi threading Sending unit and receiving unit are completely independent from each other In sending unit first data to transmit is determined which is comprised from unique HOSTNAME and PORT number of the Robot As stated earlier unique HOSTNAME is used to identify the Robot and PORT number for connection establishment which will be explained later In general unique IP address is used to identify the agents However mobile agents generally do not use static IP addresses which means IP addresses can change in every reconnection to the network Hence using IP addresses would require some adaptive techniques to handle the changing IP s Instead we decided to use HOSTNAME which is a label assigned to a device connected to a computer network for identification Ali Osman ISIK Master of Science Thesis 3 1 Design of the DRL 41 As a requirement of the Presentation Layer we have to provide common representation of data to be transferred between application entities In this version of the library Practical C Sockets interface allows us to send characters directly without any encoding In the future if lower levels of the DRL are implemented it might not be possible to use characters that easi
22. to ALPHA5 without establishing a connection in the network shown in figure 4 3 At first glance the DRL does not seem to provide any function to relay messages within the network The reason is that the DRL is designed to set up connections in the network and enable easy data transfer between two communication entities From this point of view message relaying would be a complicated task for DRL However DRL allows us to call the functions of any robot by sending a single XML message If we consider message relaying as a feature of the robot rather than something that DRL provides then we should achieve message relaying by adding an extra software feature to the robot This is done by adding a simple relay function to listoffunctions h To be able to send messages by using multi hop relay first thing is to determine a path between the sender and the receiver However this should be something to be done in user level rather than losing the simplicity of the DRL In this application we assume that the path is predefined as ALPH Al gt ALPHA3 gt ALPHA4 gt ALPHA5 These paths can be found by implementing optimal path finding algorithms As the relaying function we implemented SendToOtherRobot function which is called by an XML message provided with receiver and real function information For instance to call SynchronizeTime function of ALPHA from ALPHA1 the following message should be sent to ALPHA3 lt SendToOtherRobot gt lt RECE
23. two signals Request to Send RTS and Clear to Send CTS as we explained before Association Reassociation and Disassociation The active and passive scanning process explained for ZigBee is the same for WI FI Either Ali Osman ISIK Master of Science Thesis 2 3 High level Communication Protocols 33 a station searches for a access point AP or an AP periodically sends a beacon When a station wants to join a BSS it chooses it according to signal power signal quality and network load After data is exchanged for association authentication is preceded for reliable data transmission The synchronization is again sustained by periodically received beacons from the AP Disassociation can be carried out both up on request of AP or the station itself Reassociation happens in several cases which are roaming high traffic change of the radio environment or high error rates In high traffic reassociation brings load balancing The principle of roaming is similar to handover in GSM In mobile wireless networks a client can move out of the range and should reassociate to another AP by concerning the same criteria as association The process of dynamic association and reassociation gives the network ability to cover a wide range area Security The basic security issues of Wi Fi follows WEP Wired Equivalent Privacy standards WEP applies following security services e The shared key authentication is used In the association process of a station
24. waited for two times the maximum amount of time an IP datagram might live in the internet The reason for this is that while the local side of the connection has sent an ACK in response to the other side s FIN segment it does not know that the ACK was successfully delivered As a consequence this other side might re transmit its FIN segment and this second FIN segment might be delayed in the network If the connection were allowed to move directly to the CLOSED state then another pair of application processes might come along and open the same connection and the delayed FIN segment from the earlier incarnation of the connection would immediately initiate the termination of the later incarnation of that connection Sliding Window Algorithm TCP implements an algorithm for flow control called Sliding Window The window is the maximum amount of data we can send without having to wait for ACKs In summary the operation of the algorithm is as follows 1 Transmit all the new segments in the window 2 Wait for acknowledgment s to come 3 Slide the window to the indicated position and set the window size to the value adver tised in the acknowledgment When we wait for an acknowledgment to a packet for some time and it has not arrived yet the packet is retransmitted When the acknowledgment arrives it causes the window to be repositioned and the transmission continues from the packet following the one transmitted last The sliding window se
25. 4 Handling Relayed Messages In the first section of this chapter we showed the difference of XML messages between calling setter and getter functions In order to get some data from another robot the sender should send the handler function in the XML message In the previous section we showed how to relay messages in the network by using the DRL but still we did not solve the problem of getting data from a robot that is not directly connected Since we already have the relaying function all we have to do is to find a way to keep the path information till the receiver since it will need that information to send the data back Since we are sending handler information in XML message and we have the route information before starting we can add the sender information to the handler If there is multi hop relaying then more than one sender can be added to the handler If we use the same path as the previous section then the handler item will be as follows lt HANDLER gt lt SENDER gt ALPHA3 lt SENDER gt lt SENDER gt ALPHA1 lt SENDER gt lt FUNCTION gt PrintMessage lt FUNCTION gt lt HANDLER gt Master of Science Thesis Ali Osman ISIK 68 Experimental Evaluation of Distributed Robotics Library As we see in the message the last sender is added as the first sender as a child node Note that there is no need to add ALPHAJ as a sender because ALPHAJ3 will receive the message from ALPHA connection and it will automatica
26. 6 have focused on a distributed coordination which implies that the characteristics of the group s behaviour are not managed centrally by one or few leaders but are the result of self organizing processes In their project groups of robots evolve a cooperative strategy for tasks which can be modeled as ant colonies To summarize DCSs use the coordination of agents to increase the efficiency of the control systems 1 1 Focus and Contribution of the Thesis The Distributed Robotics Group at the Delft Center for Systems and Control DCSC is planning to address many problems in distributed robotics area especially about sensing and distributed control of mobile networks In order to address these problems a testbed which consists of multiple robots and sensors is being built as shown in Figure 1 1 Currently there are 6 Robots and each of them use a netbook for communication and running complex control algorithms The testbed is surrounded by beacons for localization of the robots Since each robot is able to perform complex calculations and can be programmed as an intelligent agent the testbed is a MAS with intelligent agents and many distributed control applications can be tested In order to apply and test the distributed control techniques on this system the agents must be interconnected by a communication network Since each robot has netbooks and has wireless network adapters the communication network can be set up by a wireless ro
27. 91 197 Springer 2005 4 F Ruini Communication and distributed control in multi agent systems Preliminary model of micro unmanned aerial vehicle mav swarms 2008 5 M Koes I Nourbakhsh and K Sycara Constraint optimization coordination architec ture for search and rescue robotics in Proceedings of the IEEE International Conference on Robotics and Automation ICRA 2006 pp 3977 3982 May 2006 6 V Trianni and M Dorigo Self organisation and communication in groups of simulated and physical robots Biological Cybernetics vol 95 pp 213 231 August 2006 7 V Juli n and C Carrascosa Physical agents in Issues in Multi Agent Systems Whitestein Series in Software Agent Technologies and Autonomic Computing pp 117 143 Birkh user Basel 2008 8 F Bullo J Cortes and S Martinez Distributed Control of Robotic Networks A Math ematical Approach to Motion Coordination Algorithms Princeton NJ USA Princeton University Press 2009 9 R Hekmat Ad hoc Networks Fundamental Properties and Network Topologies Springer 2006 Master of Science Thesis Ali Osman ISIK 78 Bibliography 10 D Wetteroth OSI Reference Model for Telecommunications McGraw Hill Professional 2001 11 ITU T Recommendation X 200 Information technology open systems interconnec tion basic reference model The basic model 1994 12 A S Tanenbaum Computer networks 2nd edition Upper
28. DRL Execution Master of Science Thesis 3 3 Execution of the DRL 55 4 10 11 12 13 14 15 16 17 18 19 In the constructor of Robot1 ListenConnectionRequests thread started and Robot1 started to listen connection requests by creating a server on PORT1 In the constructor of Robot1 Advertise thread started and advertisements started to be sent each second A DistributedRoboticsLib object Robot2 is created with the H2 and PORT2 The function handlers of Robot2 are added by calling CreateMessageHandler function In the constructor of Robot2 Listen ToAdvertisements thread started and Robot2 started to listen advertisements on ADV_ PORT In the constructor of Robot2 ListenConnectionRequests thread started and Robot2 started to listen connection requests by creating a server on PORT2 In the constructor of Robot2 Advertise thread started and advertisements started to be sent each second Listen ToAdvertisements thread of Robot1 detected advertisement of Robot2 and created a DistributedRobotCommLib object Robot2Comm with H2 and PORT2 Robot1 sent a connection request to H2 and started waiting new port number ListenConnectionRequests thread of Robot2 detected a connection request from H1 and created a DistributedRobotCommLib object Robot1Comm with H1 and PORT1 since H1 was not on the list Then it randomized a new port number and sent it to Robot1 Then it called StartCommunicat
29. FIN_WAIT_2 CLOSING L7 ACK Timeout after two segment lifetimes TIME_WAIT CLOSED Figure 2 5 TCP State Transition Diagram TCP State Transition Diagram The two transitions leading to the ESTABLISHED state correspond to the opening of a connection and the two transitions leading from the ESTABLISHED state are for the termination of a connection The ESTABLISHED state is where data transfer can occur between the two ends in both the directions If a connection is in the LISTEN state and a SYN segment arrives the connection makes a transition to the SYN_RCVD state and takes the action of replying with an ACK SYN segment The client does an active open which causes its end of the connection to send a SYN segment to the server and to move to the SYN_ SENT state The arrival of the SYN ACK segment causes the client to move to the ESTABLISHED state and to send an ACK back to Master of Science Thesis Ali Osman ISIK 22 Communication in Networks the server When this ACK arrives the server finally moves to the ESTABLISHED state In other words we have just traced the THREE WAY HANDSHAKE In the process of terminating a connection the important thing to keep in mind is that the application process on both sides of the connection must independently close its half of the connection The main thing to recognize about connection teardown is that a connection in the TIME_ WAIT state cannot move to the CLOSED state until it has
30. Go to thesis_library source applications SimpleRobotLibrary folder 3 Add header and cpp files of all the entities in your robot and add interface functions dec larations to ListOfFunctions h and definitions to ListOfFunctions cpp Note that there are some useful functions already provided with the library 4 Make the proper changes to the CMakeLists txt file of the folder Now your robot is ready to use the library 5 Go to thesis_library source applications folder 6 Add a folder for your application and make proper changes to the CMakeLists txt 7 Add your application s header and cpp files and create a CMakeLists txt file for your application 8 It is possible to add multiple application in the same way A 2 Compiling the library and applications After adding the robot files and application files go back to the home folder of library 7 thesis_ library and follow these steps Master of Science Thesis Ali Osman ISIK 76 User Manual 1 Run prepare file in order to create folders and makefiles If your system is mission any of the packages needed you will get warnings to install them 2 Run make command in order to create binary files Note that different makefiles are created for each folder to compile them separately 3 Go to thesis_library binaries applications folder where you can find the executable file 4 You can copy and use the executable on different systems
31. IVER gt ALPHA4 lt RECEIVER gt lt FUNCTION gt lt SynchronizeTime gt lt MSEC gt 101010010 lt MSEC gt lt SynchronizeTime gt lt FUNCTION gt lt SendToOtherRobot gt As we see we are adding two child items one is receiver and one is the real function to be called When this message is sent to ALPHA3 SendToOtherRobot function is called from listoffunctions h as expected In this function the file is parsed and the sub document inside the lt FUNCTION gt item is sent to the receiver At the receiver side the function in the sub document is called which is intended Multi hop relay communication is also possible by using nested SendToOtherRobot tags provided with the next receiver information After providing relaying function we tested this function by calling TurnOffMonitor function of ALPHAS from ALPHA1 After setting up the connections shown in figure 4 3 we send the XML message to from ALPHA to ALPHAS The outputs including sent and received messages are as follows Output of ALPHA1 Master of Science Thesis Ali Osman ISIK 66 Experimental Evaluation of Distributed Robotics Library Message sent to ALPHA3 lt SendToOtherRobot gt lt RECEIVER gt ALPHA4 lt RECEIVER gt lt FUNCTION gt lt SendToOtherRobot gt lt RECEIVER gt ALPHA5 lt RECEIVER gt lt FUNCTION gt lt TurnOffMonitor gt lt FUNCTION gt lt SendToOtherRobot gt lt FUNCTION gt lt SendToOtherRobot gt Output of ALPHAS Mes
32. ML parser class is utilized for encoding and decoding of messages The DRL is using multi threading in order to run different processes independent of each other This parallel way of execution increases response time of the robots and reduces the propagation delay of messages trading off with the system resources In this version of the library we assumed that each robot has enough resources to handle the communications and the library is designed for maximum response time and minimum propagation delay For testing the DRL five Asus Eee PC s are used as robots Although they do not have real actuation capabilities they have internal data to exchange and some attributes to set which are enough for testing In order to use the DRL one should add listoffunctions to the library which plays interface role between robot s real functions and common function representation of the library The functions in listoffunctions h file represents the attributes of the robot to be used in the network We added some functions for testing and in our routine function tests all the functions worked properly As we said the DRL s perception of the robot is limited to listoffunctions h Hence it is possible to add some abstract attributes to the robot by adding functions to this list In Chapter 4 we add one abstract attribute which is relaying messages Since multi hop communication is important to balance propagation delay and network resour
33. Once a client is connected to the server it creates Handle TCP Client thread and sets is connected to 1 StartCommunicationAsClient This function creates a new TCP socket to connect to the TCP server created in the above function Once it is connected it creates Handle TCP Client thread and sets is _ connected to 1 HandleTCP Client In this thread the socket which was created in the functions above is waiting for a message Once it receives a message it retrieves function name to be called from the message and searches function list of the Robot for a match to the function name Once it has a match it calls the function with the function pointer Then it goes back to message listening mode Once it receives an empty message which means the other party somehow stopped the communication it removes this RobotComm from the robot_comm_ list of the Robot Once an exception is caught from the socket it means that there is something wrong with the connection Then is connected is set to 1 and HandleTCPClient thread is killed The message format sent and received by this function looks like lt ChangeLocation gt lt X gt 10 lt X gt lt Y gt 15 lt Y gt lt ChangeLocation gt Note that ChangeLocation is a setter function Hence there will be no data to be returned back However when the function is a getter function or any function that returns data the XML message should include handler information of the sender Hand
34. Saddle River NJ USA Prentice Hall Inc 1988 13 W R Stevens TCP IP illustrated vol 1 the protocols Boston MA USA Addison Wesley Longman Publishing Co Inc 1993 14 H Labiod A Hossam and C D Santis Wi Fi Bluetooth Zigbee and WiMaz Secaucus NJ USA Springer Verlag New York Inc 2007 15 P Kinney Zigbee technology Wireless control that simply works www zigbee org resources Oct 2003 Whitepaper 16 D Steinberg and S Cheshire Zero Configuration Networking The Definitive Guide O Reilly Media Inc 2005 17 S J Donahoo and K L Calvert TCP IP Sockets in C Practical Guide for Program mers San Francisco CA USA Morgan Kaufmann Publishers Inc 1st ed 2002 18 E T Ray Learning XML O Reilly Media 2001 19 First Objective Software Inc Cmarkup c xml parser http www firstobject com 1999 2009 20 A Cedilnik B Hoffman B King and A N Ken Martin Cmake http www cmake org 2000 2010 21 Cygnus Solutions Cygwin http www cygwin com 1995 2010 Ali Osman ISIK Master of Science Thesis Glossary List of Acronyms DRL MAS OSI DCS DRS DCSC UDP TCP ASCII XML ISO FHSS DSSS OFDM HR DSSS WLAN WPAN DLL LLC Master of Science Thesis Distributed Robotics Library Multi Agent System Open Systems Interconnection Distributed Control System Distributed Robotics System Delft Center for Systems and Control User Data
35. Software Infrastructure for Communication in Distributed Robotics Systems Ali Osman ISIK DistributedRoboticsLib Delft D e ft University of I Technology Delft Center for Systems and Control Software Infrastructure for Communication in Distributed Robotics Systems MASTER OF SCIENCE THESIS For the degree of Master of Science in Embedded Systems at Delft University of Technology Ali Osman ISIK November 25 2010 Supervisors Prof dr R Babuska Dr G A D Lopes A Simonetto Faculty of Electrical Eng Mathematics and Computer Science Delft University of Technology Delft e t University of Technology Copyright Delft Center for Systems and Control All rights reserved Abstract Distributed Control Systems DCSs are being more popular in process control autonomous systems and safety critical systems because of several advantages over centrally controlled systems The Distributed Robotics Group at the Delft Center for Systems and Control DCSC is currently researching on distributed control of mobile networks In order to to apply distributed control on a mobile network agents in the network must communicate to each other Communication libraries that are created for Multi Agent Systems MASs and Distributed Robotics System DRS are generally application specific and do not address many systems with different capabilities The purpose of this project is to design and implement a software communicat
36. Time function and since it is a getter function an XML message is prepared to send the time back to ALPHA1 In the output of ALPHA1 we see that after getting the information with the handler function PrintMessage it prints the clock of ALPHA5 Since ListOfFunctions h is created by user a user can use different protocols to relay the messages or to handle the relayed messages In this section and previous section we gave our own protocols to handle message relaying and we provided some functions as an extension to the library Ali Osman ISIK Master of Science Thesis Chapter 5 Conclusions and Future Work 5 1 Summary and Conclusions The goal of this graduation project was to develop a communication library to be used by dis tributed robotics networks The main requirements of the library were portability to different operating systems to provide automatic identification in the network to provide fast and simple ways of communication to enable parallel communications to provide common struc ture for handling messages and to be open to improvements While designing the Distributed Robotics Library DRL these requirements were taken into consideration and they were tried to be provided as much as possible A communication library design requires set of protocols from physical connection to the user These set of protocols are layered and structured in the 7 Layer OSI Reference Model This model is quite abstract and any co
37. a bluetooth device such as standby page inquiry connection and park Except inquiry these states are managed by LMP Here we will explain some of the LMP procedures in detail LMP_Sniff_Req LMP_Unsniff_Req In the sniff mode a node gets into a slower cycle and saves energy A master or slave can request for a sniff mode If a link is in the sniff mode only the master can begin a transmission LMP_Host_Connection_Req LMP_Setup_Complete A connection can be established by sending a paging message on the page scan channel If the master knows the address of the slave connection is established If not the following procedure is followed e If the master does not know the address of the slave to communicate it is in inquiry state and sends a signal if the node needs a connection Then the node is in inquiry scan Master of Science Thesis Ali Osman ISIK 28 Communication in Networks e The slave goes into inquiry response substate The response consists of MAC address of the slave After the response slave goes into page scan state and then standby state to listen periodically for the messages After master get the response it goes to page state and stores the address of the slave e If the master wants to establish a link with this slave it returns a response message to the slave If the slave sees a message targeting to itself it goes into Slave response substate and sends its access code to the master e When master gets sla
38. ag is on Once a connection is established the Ack flag is always on The Acknowledgement SequenceNum and Advertised Window fields are all in volved in TCP s sliding window algorithm The Acknowledgement and Advertised Win dow fields carry information about the flow of data going in the other direction In TCP s sliding window algorithm the receiver advertises a window size to the sender This is done using the AdvertisedWindow field The sender is then limited to having no more than a value of Advertised Window bytes of un acknowledged data at any given time The receiver sets a suitable value for the Advertised Window based on the amount of memory allocated to the connection for the purpose of buffering data The header length gives the length of the header in 32 bit words Master of Science Thesis Ali Osman ISIK 20 Communication in Networks 7 f 6 4 bit header length 15 16 31 16 bit source port number 16 bit destination port number 1 32 bit sequence number 32 bit acknowledgment number 120 bytes reserved 16 bit window size 6 bits 16 bit TCP checksum 16 bit urgent pointer ae AE ve E options if any v data if any Z Figure 2 4 TCP Header e The 6 bit Flags field is used to relay control information between TCP peers The possible flags include SYN FIN RESET PUSH URG and ACK The SYN and Fin flags are used when establis
39. an ISIK Master of Science Thesis 4 3 Relaying Messages 63 lt SynchronizeTime gt ALPHA4 Figure 4 3 Topology of the Network used in Section 2 3 and 4 Since SynchronizeTime function is a setter function we just send the clock of the sender in MSEC tag After establishing the connections each robot waits for 5 seconds and they start to send the XML message to the robots that they are connected to They send the same message in each second At the 15th second we add a noise of 10 seconds to the clock of ALPHAS and at the 25th second we add the same amount of noise to the clock of ALPHA to see the responses of robots to the noise In figure 4 4 we see the graph of clocks for 45 seconds Note that clocks are shifted towards 0 with respect to the smallest clock Before the robots start the clock synchronization the first 5 seconds the clocks of the robots increase linearly After the 5th second all the clocks converge to the same value Since ALPHAJ3 is connected with only one robot its settling time is longer than the others At the 15th second and 25th second when we add 10 seconds noise to the clocks of ADPHA5 and ALPHA3 respectively there are jumps in corresponding robots clocks As we see from the graph while the clock of ALPHA3 is settling fast in 2 seconds it takes about 10 seconds for the clock of ALPHA5 to settle As expected when number of robot s connections increase the settling time of the clock decre
40. ase In this type of an application trade off between saving resources and settling time should be taken into account 4 3 Relaying Messages Message relaying is essential in a network Sometimes in order to save the resources of the network total number of connections can be limited In this case multi hop communication is used between directly unconnected agents and message must be relayed by the intermediate nodes In Bluetooth scatternets some nodes are used to connect two different piconets These intermediate nodes must have message relaying attribute in order to provide communication Master of Science Thesis Ali Osman ISIK 64 Experimental Evaluation of Distributed Robotics Library o Clock vs Time Graph ofthe Robots Clock milliseconds 0 i i 0 5 10 15 20 25 30 35 40 4 Time seconds Figure 4 4 Clocks of The Robots In Time Ali Osman ISIK Master of Science Thesis 4 3 Relaying Messages 65 between the piconets Although in this version of the library a robot cannot be connected to two different Wi Fi networks so cannot be intermediate node between networks in future protocol free versions of library Bluetooth kind of technologies can be used and a robot can be used to transfer messages between networks to increase the range of the communication In this case message relaying would also be essential for the intermediate robot In this application we will answer how can ALPHA1 send a message
41. aves and topology of the network It also connects different PANs by applying routing algorithms Master of Science Thesis Ali Osman ISIK 32 Communication in Networks 2 3 3 Wi Fi Like ZigBee being developed on top of IEEE 802 15 4 specification Wi Fi is a trademark of the Wi Fi Alliance developed on top of IEEE 802 11 WLAN standards Wi Fi supports three different operational modes In the infrastructure mode the wireless network consists of at least one access point AP connected to a fixed network and client stations connected to the access point This mode is based on cellular architecture each cell is called BSS Basic Service Set consisting of one AP Each BSS is executing same MAC protocol for multiple access within the range Second mode is ad hoc mode often referred as peer to peer mode In this mode there is no need to an AP and each station can establish a communication with any other station in the cell These cells are called IBBS Independent Basic Service Set This mode allows to create quickly and simply a wireless network where there is not fixed infrastructure or where such an infrastructure is not necessary for the required services hotel room conference centers or airport 14 The third mode is Mesh Configuration which is the hybrid configuration of both infrastructure mode and ad hoc mode Physical Layer We have seen two different physical layers before Frequency Hopping Spread Spectrum FHSS based physical la
42. cal connection such as Ethernet wireless etc In the current version of the library we use Wi Fi and IPv4 standards which restrict the library For local networking services of TCP IP protocol are more than necessary and it requires a strong infrastructure In Chapter 2 we explained Zero Configuration Networking which gives a guide to design Ali Osman ISIK Master of Science Thesis 5 2 Future Work 73 local networking protocols Similar set of protocols can be designed for lower layers of the DRL to make it more independent and more suitable for the intended applications e For supervision of the network a special node can be introduced In order to enable this node to control connectivity of the network simple special functions can be added to the listoffunctions of the robots by default These functions will allow the supervisor node to use the library functions of other robots and the supervisor node will be able to establish a communication between two robots Master of Science Thesis Ali Osman ISIK 74 Conclusions and Future Work Ali Osman ISIK Master of Science Thesis Appendix A User Manual A 1 Creating your own application The Distributed Robotics Library DRL is packaged in a single file called thesis_ library and provided as a zip file called thesis_library zip After installing the thesis_library zip on your system you need to follow these steps 1 Unzip the file to your system 2
43. ce Object ZDO and addresses three main items service discovery security and binding 14 ZigBee Device Object In ZigBee networks data is abstracted as much as possible into Key Value Pairs KVPs The coordinator has a master look up table known as a binding table that lists which nodes are interested in a particular KVP 15 The process of securely connecting endpoints of the KVPs is known as binding It is simply the ability to match two or more devices together based on their services In a ZigBee application KVP clusters are grouped to achieve a particular task These groups are called profiles These are almost the same as the Bluetooth profiles If you cannot find an appropriate profile for your application you can write your own profile Service discovery is facilitating the procedure for locating some services via their profile identifiers Although some security issues are mentioned in IEEE 802 15 4 ZDO complements it for security At each layer security is applied by encrypting the frame and authenticating when it is received The authentication is done by security services at ZDO To authenticate and derive a shared secret link establishment link key is used After security handshaking of two devices a secure transmission can be achieved Another component of ZDO is network manager which is implemented in the coordinator and it decides which PAN to connect In the creation of a new PANSs it decides the coordinator and sl
44. ces message relaying is vital for a robot By adding relay function and by calling this function in a proper way the DRL provides message relaying in the network In conclusion the DRL is a simple communication library easy and fast to use for simple applications However the significance of the DRL is that it is open to application specific improvements and it provides a design path for an advanced communication library 5 2 Future Work e The maximum message size that the DRL can send is limited with the limit in the Berkeley sockets application programming interface API which is used by the Prac tical C Sockets Although this limit is not small 2 KB for TCP 30 KB for UDP to send simple messages it may be small for sending different multimedia contents In order to send large bulks of data a protocol can be introduced and a new function can be added to the library e In DistributedRoboticsLib there is a pointer info which is passed to the object by the user This pointer comprised of some information about the robot In this version of the library we pass empty string for the information of the robot since we assume that we know the other robot s functions As an improvement this information can be filled with the attributes of the robot and this information can be exchanged once in a while to be aware of the capabilities of the network Across the planet communication is supplied by TCP IP independent of the physi
45. changing clock values In Section 4 3 we show how the DRL provides message relaying in the network and in the last Section we show the way of handling the relayed message by the sender 4 1 Testing Basic Functions Before invoking the basic functions of the library we create the library object and add function handlers to the object which are working as interfaces between real functionalities of the robot Master of Science Thesis Ali Osman ISIK 58 Experimental Evaluation of Distributed Robotics Library and the library The piece of code is as follows DistributedRoboticsLib robot hostname port info VirtualRobot my_robot new VirtualRobot 5 10 robot CreateMessageHandler MoveFirstLeg amp MoveFirstLeg my_robot robot CreateMessageHandler GetTime amp GetTime NULL robot CreateMessageHandler PrintMessage amp PrintMessage NULL After creating the library object with a proper hostname and a random port we create a VirtualRobot object to represent an entity inside the robot It can be a sensor an arm a clock etc In our case we implemented a VirtualRobot class inside the PC which has a leg Then we add the message handlers to the robot As we explained earlier the first input of the message handler is the tag name to call the function The second input is the pointer to the function to be handled which is inside listoffunctions h file The third input is the pointer to the entity inside the robot
46. d then there is a inactive period for the device until the next beacon The systems using superframe structure employs slotted CSMA CA for multiple access and they can sleep during the inactive period while the other systems employ unslotted CSMA CA and they should always be active These are used in CAP For low latency required applications beacon oriented systems can use GTSs for transmission In these time slots CSMA CA is not used and channel access is guaranteed Transfer Models There are three different data transfer model coordinator to device device to coordinator device to device In beacon enabled mode data transfer from coordinator to device starts with a beacon indicates the pending data Then device sends data request and coordinator sends acknowledgment message followed by data The device can send optional ack message From device to coordinator device listens for the beacon When it finds the beacon it sends the data The coordinator can send optional ack message to the device From device to device the transmission is done by simply sending data to the targeted peer and getting optional ack message back In all those data transfers channel is accessed by slotted CSMA CA In non beacon mode all transfer models are the same without the beaconing In this case channel is accessed by unslotted CSMA CA Association Deassociation with PANs Association starts with an active or passive scan of a device A device can either send
47. ddress when there is no DHCP server Second layer is automatic resolution and distribution of computer Host names when there is no DNS server Third layer is automatic discovery of network services IP Addresses without DHCP Each device on a network needs at least one unique IP address There are three ways of obtaining an IP address First one is manual addressing by a network administrator However with this technique one should be expert on networks for proper configuration Second method is having a DHCP server assigning IP addresses to the nodes When there is no DHCP server IP addresses can be self assigned in other words each device is responsible for choosing its own IP address which is called link local address configuration and zeroconf protocol is using this technique In link local addressing there are 65024 addresses available IP obtaining procedure is as follows e New host randomly selects IP address U out of 65024 available addresses e Broadcasts a probe message Who is using address U For probing Address Resolution Ali Osman ISIK Master of Science Thesis 2 3 High level Communication Protocols 35 Protocol ARP is used In this protocol you broadcast an ARP message to reach desired IP e If you get an ARP reply it means another host is claiming the same address In this case the host chooses another address e If there is no ARP replies it means that address U is available e However there can be message l
48. deling of Communication Libraries 17 The lowest IP address is 0 0 0 0 and highest is 255 255 255 255 In this addressing all the hosts in a network have the same network number However with the increase in the network size this IP addressing can cause some problems Instead in 1993 the network classes explained above were replaced by Classless Inter Domain Routing CIDR which allows redivision of classes There are a lot of issues in IP addressing which will not be discussed here e Options Options field can be up to 40 bytes Some of the IP options are as follows Security option tells the level of secrecy of the datagram Strict source routing option gives a complete path from source to destination Loose source routing gives a list of routers not to be missed Record route allows each router to append its IP address Timestamp option allows each router to append its IP address and timestamp In IP at fragmentation stage total length MF fragment offset and header checksum fields are computed and a header is put to all fragments of the IP datagram In reassembly stage fragments are assembled by looking at the values in their headers and datagram is passed to the next level for further usage Layer 4 Transport Layer The Transport Layer is the fourth level in OSI model It is the middle layer of the OSI model and Transport Layer data stream is a logical connection between end points The upper layers of the
49. e 1 is a connectionless service in which data is simply transmitted regardless of whether destination station accepted the packet It allows multicast transmission e Type 2 is connection oriented service First a logical connection is established between two stations The connection is maintained and terminated by using special packets that can be understood by stations Sequence numbers are added to the packets for ordered transmission Error control and flow control are provided but this service does not allow broadcast or multicast addresses e Type 3 is acknowledged connectionless service for point to point communication only The LLC header consists of 4 fields e 8 bit DSAP Destination Service Access Point is the SAP Service Access Point where the packet is intended e 8 bit SSAP Source Service Access Point is the SAP Service Access Point where the packet is originated e 8 bit control field for use in auxiliary functions such as flow control e Information field that carries LCC information IEEE 802 2 is independent of the particular medium and commonly used Ali Osman ISIK Master of Science Thesis 2 2 Modeling of Communication Libraries 15 Layer 3 Network Layer The Network Layer is the third level in OSI model We can list the functions of this layer as follows e The Network Layer is responsible for the controlling the delivery of the packets from source to destination by attaching the packets its own contro
50. e agent is able to do should be defined as a function which has a fixed structure So every message that is sent to robot becomes a function call The Robot keeps pointers to these functions in a list Whenever a message comes and is parsed by RobotComm the pointer to this function is found in the Robot and function is called with the provided parameters in the message As a simple example shown in Figure 3 8 assume that Robot2 wants to change the location of Robot1 to 10 15 by using the ChangeLocation function already exists in the list of Robot1 s functions The XML message that is sent to Robot1 looks like lt ChangeLocation gt Master of Science Thesis Ali Osman ISIK 46 Implementation of the Distributed Robotics Library Is Robot2Comm Add Robot2Comm in the list Create New Port to Comm Aicatic a as Robot2Comm Robot2Comm sta Figure 3 7 Flowchart of Connection Request Process on Robotl Ali Osman ISIK Master of Science Thesis 3 2 Main Functions of Distributed Robotics Library 47 lt X gt 10 lt X gt lt Y gt 15 lt Y gt lt ChangeLocation gt After Robot1Comm of Robot2 sends this message to Robot2Comm of Robotl which was already listening to coming messages it receives the message parses it finds the Change Location function of Robot1 and calls the function ChangeLocation function includes the actions that Agentl has to do to change the physical location to 10 15 Thus
51. e those set of rules and they exist at many levels from physical connection to end to end services The International Organization for Stan dardization ISO developed an abstract description for layered communications and network protocol design in 1974 which is called OSI Reference Model This model can be applied to any communication system It standardizes network architecture and encourages vendors to develop network equipments avoiding proprietary design Basically it divides network architecture into seven layers which from top to bottom from 7 to 1 are the Application Presentation Session Transport Network Data Link and Physical Layers as shown in Fig ure 2 2 A layer is a collection of conceptually similar functions that provide services to the layer above it and receives service from the layer below it 10 First three layers are the interface layers and the last four layers pertain to end to end functions such as user appli cation user services user interface session establishment Two instances of the same layer are connected by a protocol connection on that layer Therefore in order to understand a communication library first we should be aware of the properties of its layer Layer 1 Physical Layer The Physical layer is the first and lowest layer in OSI model The Physical Layer provides the mechanical electrical functional and procedural means to activate maintain and de activate physical connections for bit transmiss
52. ement and Robot2 adds RobotiComm to its own list At t 1 Robot2 sends the advertisement Since Robot2Comm is already in the list it is informed to stay alive Master of Science Thesis Ali Osman ISIK 42 Implementation of the Distributed Robotics Library Advertise t 0 Robot2 arrives Robot2Comm is created Advertise Robot1 Robot2Comm soe is created t 1 Robot2 Robot2Comm Robot1Comm Stay aliv Figure 3 4 Advertisement process between two agents Ali Osman ISIK Master of Science Thesis 3 1 Design of the DRL 43 Is Robot2Comm in the list Figure 3 5 Flow Chart of Advertisement Process on Robotl Master of Science Thesis Ali Osman ISIK 44 Implementation of the Distributed Robotics Library Connection Request Accept Process Connection request accept process is used to establish communication between two Robot entities A Robot in the network can send a connection request to a specific IP address The connection request process starts with finding the RobotComm with the IP address If there is no RobotComm with that IP address or there is already a connection with that RobotComm then the connection request is ignored Otherwise the Robot creates a TCP socket to send a connection request to the target Robot If the TCP socket cannot be created for any reason the process is repeated until a the connection request is sent If the socket is successfully created then the advertisement info
53. er Unlike the centrally controlled systems where the central controller is fully interacted with the agents DCSs are network centric systems and each agent requires only partial interaction with the other agents Since the knowledge is distributed in the network simpler architectures are enough for DCSs and agents have lower resource requirements than a central controller However each agent must have advanced communication attributes and agents should be interconnected by a communication network DCSs have significant advantages over centrally controlled systems Since the whole system is not critically relying on any individual the system is more robust and fault tolerant In centralized systems the knowledge is collected in one individual and the whole system is vulnerable against any attack on the knowledge in that individual Distributed knowledge in DCSs makes the system less vulnerable against any attack that is trying to compromise the security of the system Therefore nowadays DCSs are being applied more often in process control autonomous systems and safety critical systems where control needs to change to cope with fault appearance or other process disturbance 1 There are different DCS models have been proposed for different applications For exam ple Galdun and Tamac 2 propose a DCS model consisting of 4 levels technological level supervisory level information level and management level By using supervisory level and informa
54. esign of the DRL follows the 7 layer Open Systems Interconnection OSI model with some slight differences Before explaining the design of the DRL layer by layer we will give some basic information about the design The DRL is implemented in C programming language and consists of two main classes e DistributedRoboticsLib class takes care of the identification of agents establishment and maintenance of the connections in the network e DistributedRobotCommLib class takes care of the private communication between two specific agents In Figure 3 1 we give a diagram that shows the ownership of the objects that are used in Figure 3 2 As we see Robot1 Robot2 and Robot3 are objects of DistributedRoboticsLib and RobotComm s are objects of DistributedRobotCommLib From here on we will denote the objects of DistributedRoboticsLib as Robot and the objects of DistributedRobotCommLib as RobotComm Master of Science Thesis Ali Osman ISIK 38 Implementation of the Distributed Robotics Library DistributedRoboticsLib DistributedRobotCommLib gt Class Class Robot2Comm Robot1Comm Robot1Corqm Robot1 Robot2 Robot3 Robot3Comm Robot3Comm Robot2Comm Figure 3 1 Relationships of the objects with the DRL Classes In Figure 3 2 communication relationships of objects in a network consisting of 3 agents are shown If there are n agents in the network then there will be n Robots 1 for each agent However for each Robot n 1 R
55. execution of the DRL between two agents UML Sequence diagram is utilized to give a better understanding The acronyms used in the Figure 3 10 are as follows H1 Hostname of Robotl P1 Port number of Robotl Robot1 DistributedRoboticsLib with H1 and P1 Robot2 DistributedRoboticsLib with H2 and P2 R1 SCR Robot1 SendConnectionRequest H2 R1 SXML Robot1 SendXMLString H2 R1 A Robot1 Advertise R1 LCR Robot1 ListenConnectionRequests R1 LA Robot1 ListenToAdvertisements Robot2Comm Robot1 s DistributedRobotCommLib with H2 and P2 R2C SCAC Robot2Comm StartCommunicationAsClient R2C HTCPC Robot2Comm HandleTCPClient MoveLeg A function in the ListOfFunctions h file of Robot2 Note that the acronyms used for the functions of Robot2 are likewise In the following steps written flow of the DRL is given 1 2 3 A DistributedRoboticsLib object Robot is created with the H1 and PORT1 The function handlers of Robot1 are added by calling CreateMessageHandler function In the constructor of Robot1 ListenToAdvertisements thread started and Robot1 started to listen advertisements on ADV_ PORT Master of Science Thesis Ali Osman ISIK 54 Implementation of the Distributed Robotics Library Ali Osman ISIK IRIC SCAS IRZA IR2 LA eeh gl RA MoveLeg IRZ LCR EH IRoboti Comm IR2C HTCPC IRobot2Comm Figure 3 10 UML Sequence Diagram of the
56. f between the other two parameters Normally we do not want the network size too small and minimum Ali Osman ISIK Master of Science Thesis 2 2 Modeling of Communication Libraries 13 frame size too high However with the increase of network size minimum frame size should be increased and vice versa In 10 mbps Ethernet IEEE 802 3 worst case collision detection time is 51 2 usec and network diameter is around 2500 meters So minimum frame size is 512 bits which is 64 bytes Randomization of the waiting time after collision this algorithm works as follows e After the first collision stations waits either 0 or 1 slot time If each one picks the same number they will collide again e After the second collision each station picks 0 1 2 3 at random and waits the that much time slots e After ith collision each station picks a random number between 0 and 2 1 and waits that much time slots e After 10th collision maximum interval is reached and it is 0 1023 e After 16th collision the controller reports a failure and recovery is left to higher levels e One slot time is chosen as the worst case collision detection time which is 51 2 usec in case of IEEE 802 3 Because of some problems will be presented next part CSMA CD break down in wireless networks So for wireless networks another CSMA method which is Carrier Sense Multiple Access With Collision Avoidance CSMA CA is presented Carrier Sense Multiple Access with C
57. follows Connection Request sent to ALPHA2 and ALPHA5 Ali Osman ISIK Master of Science Thesis 4 1 Testing Basic Functions 59 Connection Request Sent To ALPHA2 Client started Connection Request Sent To ALPHA5 Client started List of the robots in the network ALPHA3 Not Connected ALPHA4 Not Connected ALPHA5 Connected ALPHA2 Connected From the output we understand that both ALPHA2 and ALPHA5 accepted the connec tion request which is default in this version of the DRL and server client communication processes started in which ALPHA1 is acting as a client since it sent the connection re quests Then we prepared the XML messages to be sent and ALPHA1 sends it by using SendXMLMessage function The output which also includes the XML message sent is as follows Sending message to ALPHA2 Message sent to ALPHA2 lt MoveFirstLeg gt lt X gt 15 lt X gt lt Y gt 20 lt Y gt lt MoveFirstLeg gt Here we sent a message to ALPHA2 to call MoveFirstLeg in listoffunction h file Since this function is a setter function we did not add lt HANDLER gt item to the XML message MoveFirstLeg function simply calls the MoveFirstLeg function of the VirtualRobot which was created at the beginning with the desired parameters X Y 15 20 After calling this function as we will see in the output of ALPHA2 it gets the position of the first leg from VirtualRobot and prompts to the screen Here is the whole output
58. g protocols Although the whole chapter is given as a reference guide to networking Section 2 2 is the model that we used in our library and in the design chapter we will use some of the concepts from this section Also Zero Configuration Networking subsection gives a strong method to build lower layers of a communication system for local networks and can a path for further improvements of our communication library 2 1 Modeling of Multi Agent Systems 2 1 1 Physical Agents Physical agents are situated in an environment of the real world they are embodied 7 Since they are interacting with the physical environment they have some specific individual characteristics to deal with the physical world From a very abstract perspective the basic architecture of a physical agent should consist of three components a set of sensors a set of effectors and a cognitive capability Although today s robots are missing cognitive part Master of Science Thesis Ali Osman ISIK 6 Communication in Networks robots can be good examples of physical agents We can list some of the main requirements for modeling a physical agent e A physical agent must have access to data in the physical world Sensors are used for this purpose e In a physical distributed system there should be a common global time and all agents should synchronize their clocks with that time e Ina communication process of agents in a distributed environment time representatio
59. ge Robot5Comm Client RobotiComm Server 6 Sending XML message Figure 4 2 An illustration of a getter function call ALPHA4 Not Connected ALPHA5 Connected ALPHA2 Not Connected As we see in the output after stopping the communication the robot stays in the list However server client communication process is ended by the request 4 2 Clock Synchronization Application Clock synchronization is addressing a general problem from computer science and robotics The problem is that in a Multi Agent System MAS although all the clocks are set accurately in the beginning they may differ in time because of the clock drift Clock synchronization gives a solution to the agents for converging to a common clock In this application we synchronize internal clocks of the robots by exchanging their clock information In this application we use a predefined network topology which is given in figure 4 3 In order not to change the real clock of the PCs we added clock functions to the VirtualRobot class It has a member variable delay and it keeps the track of time by the adding this delay to the real clock We are keeping the time in milliseconds Then we added SynchronizeTime function which takes the clock of the sender as an input and setting the clock to the arithmetic mean of the sender s clock and its own clock The XML message to be sent is lt SynchronizeTime gt lt MSEC gt 101010010 lt MSEC gt Ali Osm
60. gh the UDP socket This specific IP address enables us to send the message to the whole subnet that the agent is connected Although there is a unique broadcasting IP for every network this IP works for every network The Robot sends the same message in each second to the whole network While maintaining real time existence information of the Robot in the network unreliability of UDP protocol is also solved by sending out the advertisement every second In receiving side of the application layer there is a UDP socket that is listening to ADV_ PORT When a message is detected the IP address and PORT number are parsed and the list of RobotComms is checked whether the advertising Robot already exists or not If not it means that the Robot is new in the network and a RobotComm is created for this Robot and added to the list If it already exists then corresponding RobotComm is informed about its existence not to remove itself We will explain removing conditions later in detail After the advertise ment is received and processed the Robot starts to listen advertisements again We can see the advertisement process in a network with two agents in Figure 3 4 and flow chart of the process on Robot1 in Figure 3 5 When Robot2 joins the network at t 0 it sends the ad vertisement to the whole network Robot1 receives the advertisement and adds Robot2Comm to the list Before t 1 which is 1 second after Robot1 s last advertisement Robot1 sends advertis
61. gram Protocol Transmission Control Protocol American Standard Code for Information Interchange Extensible Markup Language International Organization for Standardization Frequency Hopping Spread Spectrum Direct Sequence Spread Spectrum Orthogonal Frequency Division Multiplexing High Rate Direct Sequence Spread Spectrum Wireless Local Area Network Wireless Personal Area Network Data Link Layer Logical Link Control Ali Osman ISIK 80 Glossary MAC Media Access Control VLAN Virtual Local Area Network CSMA CD Carrier Sense Multiple Access With Collision Detection LAN Local Area Network CSMA Carrier Sense Multiple Access CSMA CA Carrier Sense Multiple Access With Collision Avoidance Ali Osman ISIK Master of Science Thesis
62. hing and terminating a TCP con nection respectively The ACK flag is set any time the Acknowledgement field is valid implying that the receiver should pay attention to it The URG flag signifies that this segment contains urgent data When this flag is set the UrgPtr field indicates where the non urgent data contained in this segment begins The PUSH flag signifies that the sender invoked the push operation which indi cates to the receiving side of TCP that it should notify the receiving process of this fact The RESET flag signifies that the receiver has become confused and so wants to abort the connection e The Checksum covers the TCP segment the TCP header and the TCP data This is a mandatory field that must be calculated by the sender and then verified by the receiver e The Option field is the maximum segment size option called the MSS Each end of the connection normally specifies this option on the first segment exchanged It specifies the maximum sized segment the sender wants to receive Ali Osman ISIK Master of Science Thesis 2 2 Modeling of Communication Libraries 21 e The data portion of the TCP segment is optional CLOSED Active open S YN Close Passive open Close SYN SYN ACK Send SYN SYN SYN ACK SYN_RCVD SYN_SENT SYN ACK ACK Glose FIN ESTABLISHED Close FIN FIN ACK CLOSE_WAIT FIN_WAIT_1 Close FIN LAST_ACK ACK
63. ion between data link entities A physical connection may involve intermediate open systems each relaying bit transmission within the physical Layer The Physical Layer entities are interconnected by means of a physical medium 11 As we stated before the physical layer is supposed to provide some services for the data link layer We can list some of the services of the physical layer as follows 10 e Providing physical interface specifications such as Master of Science Thesis Ali Osman ISIK 10 Communication in Networks The Seven Layers of OSI i User Transmit Receive i Application Layer Baa Presentation Layer Sesion Layer Transport Layer Network Layer Data Link Layer gt Physical Link gt Figure 2 2 The 7 Layer OSI Model Mechanical specifications of wires and cables to ensure data delivery at the speed required Mechanical specifications of the connectors such as type size number of pins and pin configuration Electrical characteristics such as voltage level interpretation current level mainta nence signaling type duration bandwidth and strength e Concerning with the transmission techniques such as synchronous or asynchronous serial or parallel bit by bit or symbol by symbol transmission Signaling start and stop of a transmission and controlling flow of the the transmission e Detecting fault conditions and providing them to the next level
64. ion library for DRSs not only to meet the needs of Distributed Robotics Group but also to address more general MASs The library is called the Distributed Robotics Library DRL The DRL is modeled by the 7 Layer Open Systems Interconnection OSI Reference Model Since Wi Fi standards are used for the physical transmission of data between agents and the medium access control in the network agents must be Wi Fi certified or compliant in order to use this version of the library For the network layer IPv4 for the transport control TCP and UDP protocols are used These protocols provide all the necessary services for the DRL Our DRL is divided into two classes one for general communication in the network and one for the specific communication with a robot For identification of the robots in the network an automatic advertisement broadcasting technique is used For every identified robot in the network an object of the specific communication class is created Starting and stopping these specific communications are controlled by general communication class This class is using connection request accept technique to start server client communication between the robots All these processes in the library are running in parallel in order to minimize the transmission delay and to maximize data transfer rate However resource management is not taken into consideration in this project The DRL is designed to detect the robots which leave the network o
65. ionAsServer function of the RobotiComm This function created a server on the new port and started waiting for a connection After Robot1 received the new port it called StartCommunicationAsClient function of the Robot2Comm This function created a TCP socket and connected to the TCP server created by Robot1Comm Then it created a new thread HandleTCPClient and started waiting for a message After TCP server of Robot2Comm received a connection it created a new thread Han dle TCP Client and started waiting for a message Now both parties are ready to send messages to each other Robot1 asked Robot2 to move its leg by sending a proper XML message by calling SendXMLMessage function with input H2 Robot2Comm is found from the list and SendXMLString function is called Once Handle TCP Client thread of Robot1Comm received the message it searched the function list for a match of MoveLeg When it finds the match callback function is called MoveLeg function in the ListOfFunctions h file of Robot2 is called and Robot2 moved its leg to the desired position Master of Science Thesis Ali Osman ISIK 56 Implementation of the Distributed Robotics Library Ali Osman ISIK Master of Science Thesis Chapter 4 Experimental Evaluation of Distributed Robotics Library To test the Distributed Robotics Library DRL five Asus Eee PC s are used to represent five different agents robots in a network All the PCs are running 32 bit Microsof
66. istence in the network the Robot sends the advertisement information each second to the whole network Transport Layer of Advertisement Process For the Transport Layer of advertisement process User Datagram Protocol UDP is used since it uses a simple transmission model without implicit hand shaking dialogues Without the need of the address of the receiver UDP allows us to send the advertisement information to the whole network Since we are sending the advertisement each second we need fast transmission of the data which is supplied by UDP Although UDP does not provide a reliable service loss of an advertisement is not critical in a network since it is sent each second As we explained in Chapter 2 the agents can use software interfaces which is not specified in OSI model In advertisement process and the other two processes we use a very powerful interface between Layer 4 and 5 Practical C Sockets which provides wrapper classes for a subset of the the Berkeley sockets application programming interface API TCP and UDP sockets and works on both the Unix tested under Linux RedHat 7 3 with gcc and Windows tested under Win2K with Visual C 6 0 platforms are used in our library Practical C Sockets provides a very simple C interface to sockets 17 and the whole software is freeware In high level protocols such as TCP IP it is difficult to separate the upper 3 layers of the OSI model For this reason we will call the rest of the
67. l information Remember that the Data Link Layer is responsible for the node to node transmission of the frames which serves to the Network Layer e It divides messages into packets and resizes them to the sizing requirements of the receiving networks and it creates virtual circuit environments to deliver these packets in the correct sequence e It ensures the transmission speed is in the range that the receiver can candle e It maintains quality of service and error control functions e In connection oriented networks it establishes logical connection before the packet ex change Responsible for the routing of packets from one network to the other network It is responsible for unique host addressing of the nodes The best known protocol in the Network Layer is the Internet Protocol IP In the following subsection we will briefly explain Internet Protocol Internet Protocol The Internet Protocol has the task of delivering datagrams pieces of data streams from source to destination The most used version of the Internet Protocol used in Internet is Internet Protocol Version 4 Pv4 and we will explain this version of Internet Protocol In order to understand IP first we should look into IPv4 datagram format An IP datagram consists of a header and data part The header has 20 byte fix part and optional 40 bytes The rest of the datagram is the data part In Figure 2 3 we see an IPv4 header with the name of the fields and definitions
68. lave transmit alternatively such that master uses even numbered slots and slaves use odd numbered slots Data is transmitted using packets and a packet has a size of 1 3 or 5 time slots Frequency hopping timing gives 250 260 slots for settling of radio circuits and 366 bits are used for the packet If 5 slot packets are used after settling 2781 of 3125 bits are available for the packet which is much more efficient than single slot packet Logical Links The most used two types of logical links channels can be listed as follows e Asynchronous connection less ACL This type of link is for data communication Any slave has ACL link by default and a slave can have only one ACL link This link can be used for symmetrical and asymmetrical asynchronous services For early Bluetooth technology throughput can be 723 2 Kbps in one direction and 108 433 Kbps for symmetrical flows Frames can be lost and have to be retransmitted A slave can only transmit after having received a packet from the master e Synchronous connection oriented SCO This type of link is used for voice com munications such as telephone connections This channel is allocated a fixed slot in each direction This link is symmetric and can send up to 64 Kbps in each direction Since time is critical retransmission is not possible but forward error correction FEC can be used for high reliability There can be at most 3 SCO links in a piconet Packets There are three types of packet
69. ld be provided by the user who wants to use the DRL ListOfFunctions cpp is the file that consists of function definitions which act like an interface between the DRL and functions of the agent The functions should have a fixed signature which is the same as the function pointer in message _handler structure The signature of the function is as follows void MoveFirstLeg CMarkup xml void data void S The first input is the pointer to the XML message that is sent to the RobotComm The second input is the pointer to the entity of the agent from which the real function will be called It is the third input of CreateMessageHandler function that is passed to the function Third input is a pointer that is passed to the function in Handle TCP Client function It is a pointer to the RobotComm which received the message This pointer is used to send a message back to the sender RobotComm The functions in the ListOfFunctions h should be structured to do the following tasks Ali Osman ISIK Master of Science Thesis 3 3 Execution of the DRL 53 3 3 Parse XML string and retrieve the inputs that are needed to call the function Call the function If it is a setter function do nothing If itis a getter function save the output of the function Retrieve handler information from the XML String and prepare the return message by adding the result of the function Execution of the DRL In this section we will explain the
70. ler information lets receiver know how the sender will handle the data so that the receiver can send the data back as another XML message An example of a message with handler information looks like lt GetTime gt lt HANDLER gt Master of Science Thesis Ali Osman ISIK 52 Implementation of the Distributed Robotics Library lt FUNCTION gt PrintMessage lt FUNCTION gt lt HANDLER gt lt GetTime gt In this message sender asks for the time of the receiver and tells the receiver that it will handle the returned data by printing it In this case the reciever s return message would look like this lt PrintMessage gt lt STRING gt Time lt STRING gt lt PrintMessage gt WatchDog In this thread watch_dog variable is decreased by 1 in every second When it hits to 0 this RobotComm will be removed from the robot_comm_ list of the Robot StopCommunication This function stops the communication by killing Handle TCP Client thread and deleting the port Then it sets is_ connected to 0 Note that RobotComm will still stay in the robot_comm_ list of the Robot but without connection SendXMLString This method simply sends the XML message provided by SendXMLMes sage function of the Robot The message sent by this method is handled by Handle TCP Client thread of the partner RobotComm 3 2 3 ListOfFunctions Class Although ListOfFunctions h and ListOfFunctions cpp is not a part of the DRL it is a file that shou
71. lly send the reply message back to it When the final receiver gets the XML message it parses the message and prepares a reply XML message to be sent to the first sender As an extension of the library we added PrepareReturnMessage function which takes handler sub document as an input and prepares the return message by using Send ToOtherRobot tag names Then the receiver adds data to the prepared message and sends it back To test this functionality we sent GetTime function by using the same path The outputs of the robots including the messages they sent are as follows Output of ALPHA 1 Message sent to ALPHA3 lt SendToOtherRobot gt lt RECEIVER gt ALPHA4 lt RECEIVER gt lt FUNCTION gt lt SendToOtherRobot gt lt RECEIVER gt ALPHA5 lt RECEIVER gt lt FUNCTION gt lt GetTime gt lt HANDLER gt lt SENDER gt ALPHA3 lt SENDER gt lt SENDER gt ALPHA1 lt SENDER gt lt FUNCTION gt PrintMessage lt FUNCTION gt lt HANDLER gt lt GetTime gt lt FUNCTION gt lt SendToOtherRobot gt lt FUNCTION gt lt SendToOtherRobot gt Message received from ALPHAS lt PrintMessage gt lt STRING gt Localtime is 15 14 29 lt STRING gt lt PrintMessage gt Localtime is 15 14 29 Output of ALPHAS Message sent to ALPHA4 lt SendToOtherRobot gt lt RECEIVER gt ALPHA5 lt RECEIVER gt lt FUNCTION gt lt GetTime gt lt HANDLER gt Ali Osman ISIK Master of Science Thesis 4 4 Handling Relayed Messages 69 lt
72. ly e It is responsible for quality of service requirements Middleware Layer RFCOMM protocol in this layer emulates serial and USB ports over the L2CAP protocol It supports up to 60 open ports and provides reliable data transfer simultaneous connection and flow control Service Discovery Protocol is another protocol in this layer This is the layer on top of the LMP and service discovery is done after authentication Seven messages are defined to discover and restore the information from client to server Ali Osman ISIK Master of Science Thesis 2 3 High level Communication Protocols 29 Bluetooth Profiles A profile defines a set of protocol components SDP RFCOMM necessary for the set up of correct and communicating applications 14 Since there were different groups working on specific applications and generated their own profiles there are tens of profiles Now we will give the most prominent profiles Generic Access Profile is not a real application but it provides reliable and secure links Service Discovery Profile is used to discover the services other devices have to offer With Generic Access Profile all Bluetooth devices are expected to implement this profile Serial Port Profile emulates a serial line Generic Object Exchange Profile defines the server client relationships for the data to be transferred e LAN Access Profile allows the device to connect a fixed network e Dial up Networking Profile allows
73. ly In that case best standard would be American Standard Code for Information Interchange ASCII to represent data because most of the modern character encoding schemes are based on ASCII After representation of characters a representation structure of data should also be defined in this layer We used the most prominent data encoding standard Extensible Markup Language XML standard for its simplicity and generality XML is a flexible way to create self describing data and to share both the format and the data on the World Wide Web intranets and elsewhere 18 Since XML is simple and easy to use XML based formats have become the default for most office productivity tools Note that presentation layer is common in all three of the processes After encoding the advertisement message with XML it looks like this lt HOSTNAME gt ALPHA1 lt HOSTNAME gt lt PORT gt 51000 lt PORT gt In order to create XML messages and retrieve data from XML messages easily we need a XML processor which is called XML Parser Among many freeware C XML parsers we chose CMarkup because of its simplicity ease of use and speed 19 This library provides all the functions we need to create and parse XML messages After creating the advertisement message in sending unit the robot creates a UDP socket at a preshared port number ADV_ PORT which is known by all the Robots Then Robot sends the advertisement message to the IP address 255 255 255 255 throu
74. mmunication system is modeled around this model In designing of the DRL this model is utilized Communication libraries that are used for distributed robotics are generally application spe cific and in order to get the best performance all protocols are set to be suitable to the applications On the other side the DRL is designed independent of any application For this reason in lower layers of the model it uses general protocols For physical and data link layers it is designed to use Wi Fi wireless standard Hence the agents should be Wi Fi certified in order to use the DRL The network layer also uses the most prominent protocol which is IPv4 protocol There are three main processes in the DRL which are advertisement process connection request process and server client communication process These processes have different set of rules which means the DRL uses different protocols for upper layers In network layer both TCP and UDP are used UDP is used by advertisement process for broadcasting and TCP is used by the other processes for one to one communication The DRL is implemented in C Programming Language In order to use TCP and UDP protocols with C Practical C Sockets interface class is utilized For representation of messages XML encoding is Master of Science Thesis Ali Osman ISIK 72 Conclusions and Future Work used XML is chosen because it is generally accepted widely used and easy to understand CMarkup X
75. n should be integrated to the information transmitted e In a communication process of agents there should be reliable protocols that assure the communication If the application is real time these protocols must also assure the communication with low latency e There should be a controller that assures coherent use of resources and agents should implement hard resource management e Agents should be fault tolerant and should offer a recovery mechanism e If a physical agent is mobile it should have adapting and learning mechanisms 2 1 2 Geometric Model of Multi Agent Systems MASs can be represented geometrically with the help of graphs in which agents correspond to nodes and communication links correspond to edges Basics of Graph Theory A simple graph G V consists of V a nonempty set of vertices and a set of unordered pairs of distinct elements of Y called edges Simple graphs have their limits in modeling the real world Instead we use multi graphs which consist of vertices and undirected edges between these vertices with multiple edges between pairs of vertices allowed Note that simple graphs are all multi graphs In order to represent connection to itself pseudo graph is defined which allows loops in the edge set Another type is directed graph G V consists of a set V of vertices a set of edges that are ordered pairs of elements of V The basic terminology that describes the vertices and edges of
76. n Libraries 20202004 9 2 2 1 7Layer OSI Model sisare resne o 9 22221 Werte sanane a e a E TUAS A a o alte hs 24 2 3 High level Communication Protocols 2 e 24 2 3 1 Bluetooth Protocol scos e abr a a eoo a 200000000 eee 24 PS PATA Ta Eea i e IIA A e aie See 29 223 3 WiFi o h bee 8h A o Edu Se T 32 2 3 4 Zero Configuration Networking oaoa o 33 3 Implementation of the Distributed Robotics Library 37 3 1 Design of the DRL eusi cc eee ri RR e A ES 37 3 1 1 Layers of Distributed Robotics Library 38 3 2 Main Functions of Distributed Robotics Library 47 3 2 1 DistributedRoboticsLib Class Functions 49 3 2 2 DistributedRobotCommLib Class Functions 51 3 2 3 ListOfFunctions Class 2 2 2 2 52 3 3 Execution of the DRL 02 02000002 pee 53 Master of Science Thesis Ali Osman ISIK iv Table of Contents 4 Experimental Evaluation of Distributed Robotics Library 57 4 1 Testing Basic Functions 2 00000 eee ee 57 4 2 Clock Synchronization Application 2 0 0 0 0 o e 62 4 3 Relaying Messages 63 4 4 Handling Relayed Messages aa 67 5 Conclusions and Future Work 71 5 1 Summary and Conclusions 2 0 71 5 2 Future Work 2 aa a 72 A User Manual 75 A 1 Creating your own application 0 o 75 A 2 Compiling the library and applications
77. nce Thesis 3 1 Design of the DRL 45 Robot2 wants to connect to Robot1 1 Connection Request gt Robot1 4 Send New Port 2 Request is detected 5 Robot1Comm is found 3 Robot2Comm is found J Robot2Comm Ja Robot1Comm J 6 Create a server on New Port 7 Connect to the server on Robot2Comm J Figure 3 6 Connection request process Robot is already listening messages it receives the message parses the message and applies the action in the message Then RobotComm starts waiting for another message This com munication is continuous until one side is not available outside the range shut down sleeping etc or wants to stop communication Under non availability conditions the RobotComm deletes the socket and removes itself from the RobotComm list of the Robot If one party wants to stop the communication the socket is deleted but the RobotComm is still alive for further communication requests In the DRL the handling of the received message is also included in the application layer of the process After creation of a Robot the user has to provide the pointers of agents functions to the Robot Although the method will be explained in the next chapter we give some brief information about the process In a C file associated with the DRL the agent has list of its functions and their definitions To provide a common representation of agent s actions everything th
78. obotComms will be introduced to take care of communication with the other n 1 Robots So there will be n x n 1 RobotComms in the total network 3 1 1 Layers of Distributed Robotics Library The DRL is designed to be used in networks with mobile agents Thus the physical layer transmission should be wireless For the physical layer layer 1 and the data link layer layer 2 Wi Fi standards are used Theoretically ethernet protocol with an ethernet cable can also be used but since the robots are mobile practically it is not applicable Therefore the agents which want to use the DRL should support Wi Fi technology Wi Fi provides standard for the lowest two layers of the DRL It uses WLAN IEEE 802 11 standard for the physical layer and Carrier Sense Multiple Access With Collision Avoidance CSMA CA technique for the data link layer These layers and protocols that we used are explained in Chapter 2 in detail However detailed information is not needed to understand the DRL For this version of the DRL we use IPv4 protocol for the network layer layer 3 since it is the most prominent network layer protocol and supported by the PCs that will be used by the robots However IPv4 provides more services than we need for a local network and this causes transmission of redundant data along with the core data In future more suited network layer protocol can be designed for the DRL In Figure 3 3 we see first 3 layers of the DRL with the protoc
79. of the fields follow e Version This field states which version of the protocol this datagram belongs to In IPv4 case this field should be 4 e IHL Since the length of the header is not constant IHL field tells the number of words in the header Minimum value of this field is 5 and maximum is 15 which means header should be 20 60 bytes e Type of Service The first three bits of this field are precedence field i e 111 Network Control 110 Internetwork control 001 Priority etc Next three bits are D T R flags If flag is set it means corresponding Delay Throughput Reliability property is cared more Master of Science Thesis Ali Osman ISIK 16 Communication in Networks 32 Bits Critter ririri PALTA O con an LA LEA Version IHL Type of service Total length Identification F Fragment offset Time to live Protocol Header checksum Source address Destination address A Options 0 or more words Figure 2 3 The IPv4 header e Total length This field defines the datagram size The maximum number is 65535 bytes and the minimum number is 20 bytes e Identification It allows the destination host to determine which datagram an arrived fragment belongs to The fragments of a datagram all contain the same identification value e DF If this Don t Fragment DF bit is set it means the destination host is incapable of defragmentation so it is an order not to fragment the datag
80. ollision Avoidance CSMA CA The reason why CSMA CD cannot be used in WLANs and WPAN is it is not possible listening while sending in these networks so collision detection is not possible CSMA CA is simply an improvement on CSMA because of two problem arises in wireless networks In wireless networks carrier sense range for a station is limited e Suppose that two stations which are not in the range of each other both trying to send data to a third station which is in the common range Then these two stations are hidden to each other so collision occurs This is called Hidden Node Problem e Suppose that two stations which are in the range of each other both wants to send data to different stations out of mutual range This should not cause any problem for the network However these two stations senses each others transmission and delay their transmissions since they find the channel busy This is called Exposed Node Problem In CSMA CA some improvements added to CSMA to solve these problems It is supple mented by two signals Request to Send RTS and Clear to Send CTS In this protocol the sender sends a RTS to the receiver which contains the length of the data Then receiver replies with CTS which includes the same information Upon the receipt of CTS the sender begins transmission In this protocol a station is not allowed to transmit data if it senses a CTS because it means that a station in the range is about to receive data This solve
81. ols used In the DRL there are three different communication processes For the identification of the agents within the network an automatic advertisement process is developed To establish connection between agents a connection request accept process is implemented and for the actual data transfer between agents a server client communication process is used The first 3 layers explained in the previous paragraphs are common in all these three processes Ali Osman ISIK Master of Science Thesis 3 1 Design of the DRL 39 Robot2Comm Robot1Comm Robot1 Robot3Comm Robot3Comm Robot2 RobotiComm Robot2Comm Robot3 Figure 3 2 DRL Objects of a network with three robots Figure 3 3 Protocols used in the first three layers of the DRL Master of Science Thesis Ali Osman ISIK 40 Implementation of the Distributed Robotics Library However the protocols used for the upper layers are slightly different Now we will explain these processes in detail Advertisement Process Advertisement process is used by a Robot to introduce itself to the network and identify the other Robots in the network Every Robot in the network has a unique HOSTNAME and not necessarily unique PORT number While HOSTNAME is used to identify the Robot PORT number is used to identify the communication channel that Robot is using Advertisement information consists of these two variables of the Robot In order to notify other Robots about its ex
82. or wireless networks is lognormal geometric random graph model In this model log distance path loss model is used for RF signals The probability of a link between two nodes is a function the distance between them This time the range of a node is not represented by a circle but in a more realistic way This model also shows small world property and has more realistic link distribution method thus it is the best model for wireless networks Ali Osman ISIK Master of Science Thesis 2 2 Modeling of Communication Libraries 9 In lognormal geometric random graph model two ways can be used to check 1 connectivity 1 Choosing a node at random and using a simple flooding algorithm to tag all nodes belonging to the same cluster This procedure is repeated for all untagged nodes until no untagged nodes remain in the graph If the largest cluster found in this way contains all nodes the network is 1 connected 2 The second procedure for checking 1 connectivity uses the N x N Laplacian of G By looking at the eigenvalues of laplacian one can analyze many features about the con nectivity In some cases it is easier to determine connectivity by giant component size If giant com ponent includes all of the vertices then the graph is 1 connected 2 2 Modeling of Communication Libraries 2 2 1 7 Layer OSI Model In a communication system the transmitter and receiver must agree a set of rules between them before exchanging data Protocols ar
83. oss and delay of messages Depending on the protocol that is used there can be more probing For instance IPv4 is using 4 probes e If none of the probes you get an ARP reply then use U as an IP address e Then you announce your address to the network using ARP However there is always a small probability of choosing an existing IP address In that case late conflicts happen The reason for this may be network topology changes misbehaving hosts etc Whatever the reason is the conflict has to be handled Dynamic Configuration of IPv4 Link Local Addresses RFC 3927 allows the host to handle the conflict in two ways The first way is backing down and choosing another IP If it has open TCP connections which it does not want to lose second way is sending ARP announcement asserting its own ownership of the address In second way when both of the hosts does not back down finally one of them lose open connections and back down Names without DNS IP addresses may change over time and network location and IP addresses are not human friendly form to remember or recognize devices So it is easier to use text based names instead of IP addresses In global networking domain name system DNS is used However to use DNS one should set up DNS servers which is very costly When you need names that are only valid for local links link local Multicast DNS provides a simple solution In normal DNS namespace is URLs such as http tas tudelft nl in
84. otComm and will be explained later SendXMLMessage This method is used to send an XML message to a particular host name First the corresponding RobotComm is found by searching the robot_comm_ list If there is no then message is ignored If found then the message is sent by calling SendX ML String method of RobotComm The rest will be handled by RobotComm and will be explained later CreateMessageHandler This is one of the fundamental functions of the library The function simply adds a new message_handler to the function_list It has three inputs Ali Osman ISIK Master of Science Thesis 3 2 Main Functions of Distributed Robotics Library 51 e function_name This is the tag name of the callback function In XML messages this name should be used e function pointer This is pointer to the callback function which should be in ListOf Functions h file e data This is a void pointer In case the function to be called is a function of another entity inside the agent a pointer to this entity should be passed to use its functions In this case the callback function in ListOfFunctions h is used as an interface The significance of this input will be explained better in Chapter 4 3 2 2 DistributedRobotCommLib Class Functions StartCommunicationAsServer This function creates a TCP server socket on the new gen erated port by ListenConnectionRequests function of the Robot Then it waits for a connection to the server
85. r shut down for some reason In these cases failing robots are detected and removed from the network The robots Master of Science Thesis Ali Osman ISIK which sleep or leave the network automatically introduce themselves when they are back However they cannot recover previous information in this version of the library To test the DRL five Asus Eee PC s are used to represent the robots I have successfully tested all the functions in the DRL library via a number of verification applications During the tests message relaying function is implemented and added as an extension to the library Although all the PC s are running Microsoft Windows XP the DRL was also tested a Unix environment to test its portability The current version of the DRL lays the ground towards the implementation of a powerful general purpose automatic communications library In the DRL the closest three layers to the user are implemented By implementing lower layers it can be made more efficient for distributed control applications Ali Osman ISIK Master of Science Thesis Table of Contents Acknowledgements v 1 Introduction 1 1 1 Focus and Contribution of the Thesis oaoa aa e 2 132 Thesis Outlines 0 ii lt TA dt il 2 2 Communication in Networks 5 2 1 Modeling of Multi Agent Systems o e o 5 2 1 1 Physical Agents 000000 20 ee ee 5 2 1 2 Geometric Model of Multi Agent Systems 6 2 2 Modeling of Communicatio
86. ram e MF MF stands for More Fragment All fragments except the last one have this bit set so it can be used to make sure all the fragments arrived e Fragment Offset This field tells where this fragment belongs in the datagram in multiple of 8 bytes In 13 bits maximum fragments per datagram can be 8192 So 65536 bytes can be represented which is enough when we consider the maximum length of 65535 Time to live This field limits the life time of a datagram It is a hop count field and it is decremented at every hop When it hits 0 the packet is discarded and a warning packet is sent to the sender e Protocol This field tells which protocol to be used at the transport layer after defragmentation of the fragments e Header Checksum This field only verifies the header At each hop header checksum is recomputed and compared with this field When there is a mismatch the packet is discarded Source and Destination Addresses Every host on the Internet has a unique 32 bits IP address which encodes its network number and host number Until 1993 IP addresses were divided into 5 categories A B C D E which is called classful addressing Classes A B and C have different sizes for network field and host field which can all address small medium and large networks Class D is for multicast and class E is reserved for future use Addresses are written in dotted decimal format as 4 octaves Ali Osman ISIK Master of Science Thesis 2 2 Mo
87. ranslating from service name to the IP address of the node In order to be more useful Zeroconf service type gives both information about the service and which protocol has to be used For instance _ pp encodes both printing and via Internet Printing Protocol DNS protocol family already defines a record type of SRV for service discovery In Zeroconf we have DNS pointer PTR each pointing to different SRVs By performing a PTR lookup for a name of a service you can get all instances and choose one of the services In order to retrieve the information for connecting DNS SD TXT record can be read and all information such as host name IP address or the port number that corresponding service resides etc can be reached In the existing case of a Unicast DNS server same concepts can be used hierarchically to reach entire world Ali Osman ISIK Master of Science Thesis Chapter 3 Implementation of the Distributed Robotics Library In this chapter the design and implementation of the Distributed Robotics Library DRL will be introduced In section 3 1 the design of the DRL will be explained Section 3 2 explains the main functions in the library In section 3 3 connection establishment and message transfer between two agents are explained by using a UML sequence diagram 3 1 Design of the DRL As explained earlier the DRL is implemented to set up a communication network in any distributed robotics system to allow decentralized control The d
88. raph and p is the probability of having a link edge between any two nodes Very interesting property of this model is there is a critical probability at which a giant cluster forms Above a certain value a giant component emerges and spans almost entire network In this manner this graph model is very easy to analyze When p is above that threshold the graph is assumed to be connected This model also shows small world property since any node can be connected with any other one and increase in number of nodes does not increase maximum hop count In terms of showing small world property and random distribution of nodes it can be used for wireless networks however it is not realistic since the link between any two nodes is random In order to take into account the way that radio waves propagate pathloss geometric random graph model is presented In this model a node is connected to another node if its in the range The range is a simple circle around the node Obviously this model has a better solution for RF signals but it is not realistic since the range cannot be a circle Also it does not show small world property The node and link distribution of these two models is given in Figure 2 1 9 As it is also mentioned on the figure reality is somewhere in between Random graphs Pathloss Geometric random graphs Reality is somewhere in between Figure 2 1 Comparison of Wireless Network Graph Models The most realistic model f
89. reliability and long range In all of the topologies there should be one coordinator device There are two kind of capacities of a device a full function device FFD and a reduced function device RFD FFD can be a coordinator while RFD can only be a slave or normal node Master of Science Thesis Ali Osman ISIK 30 Communication in Networks Physical Layer There are 27 channels defined in total 16 of them are on the 2450 MHz ISM band 10 on 915 MHz and 1 on 868 MHz 2450 MHz physical layer generates 250 Kbps while 915 MHz and 868 MHz generates 40 Kbps and 20 Kbps respectively Unlike Bluetooth IEEE 802 15 4 uses DSSS Direct sequence spread spectrum On the 2450 MHz band offset quadrature phase shift keying O QPSK modulation technique is used while on the other bands binary phase shift keying BPSK is used These phase shifting techniques minimize power consumption and reduce complexity Another property of physical layer is energy detection ED It can sense the energy level of each channel and it is used for measurement of link quality MAC Layer Channel Access In IEEE 802 15 4 there are two different channel access methods depending on the system if it supports beaconing or not The systems that supports beaconing use superframe structure A superframe starts with a beacon at the first time slot and followed by contention access period CAP and contention free period CFP in which guaranteed time slots GTS are available An
90. rmation is sent to the target Robot to indicate the source of connection request Then the Robot waits for data until it comes Right after its creation each Robot starts to listen connection requests by creating a TCP server socket at its specific port Whenever another Robot tries to send a connection request by creating a TCP socket with the IP and port of the listening Robot it accepts the socket and waits for the connection data In this version of the DRL there is no condition for accepting or refusing a connection request Hence every connection request is by default accepted After the client Robot sends its data as explained in the previous paragraph the data is parsed and IP address and port number of client is determined Then a randomly created port number is sent to the client Robot If there is no RobotComm with this IP address which is the situation that the server Robot did not receive client s advertisement yet the RobotComm with this IP and port is added to the list Then this RobotComm creates a TCP server socket on the new port to start the communication as a server and waits for the client to connect After the further communication is handed over the RobotComm the Robot starts to listen connection requests again After client Robot gets the data that it was waiting for the new port number the Robot Comm which is responsible for the communication with the server Robot is found and that RobotComm creates a TCP socke
91. rves the following purposes e It guarantees the reliable delivery of data e It ensures that the data is delivered in order e It enforces flow control between the sender and the receiver Layer 5 Session Layer The Session Layer is the fifth layer in OSI model The purpose of the Session Layer is to pro vide the means necessary for cooperating presentation entities to organize and to synchronize their dialogue and to manage their data exchange 11 The services of the Session Layer in a connection mode transmission are as follows Ali Osman ISIK Master of Science Thesis 2 2 Modeling of Communication Libraries 23 It establishes session connection between two presentation entities upon requests of both entities and releases connection upon request of one of the entities It provides normal data transfer and expedited data transfer between two presentation entities e It provides session connection synchronization which allows presentation entities to de fine synchronization points and reset the session connection It notifies the presentation entities about the exceptions e In connectionless mode the Session Layer only provides connectionless mode transmis sion and exception reporting In many protocols Session Layer is not given as an explicit layer like in TCP IP suite In TCP IP there is only Application Layer above Transport Layer Layer 6 Presentation Layer The Presentation Layer is the sixth layer in OSI model
92. s Ali Osman ISIK Master of Science Thesis 2 3 High level Communication Protocols 27 e Control Packets These are the packets used to carry information about clock fre quency hopping link establishment or connection requests There are 5 of them ID POLL NULL FHS and DM1 e Synchronous Packets These are the packets used in SCO links and they carry the voice There are various types such as DV HV1 HV2 HV3 e Asynchronous Packets These are the packets used in ACL links and they carry data There are various types such as DM1 DM3 DM5 DH1 DH3 DH5 for different packet sizes Error Control For error detection 2 fields can be used Error control field CRC is 16 bits long and it is at the end of the data body Error checking is done by calculating CRC pattern of the payload Header Error Check HEC is 8 bit CRC and used to check header in the same manner After detecting errors there are 3 types of Forward Error Correction FEC e In 1 3 FEC Coding each bit is repeated 3 times and decoding is done by calculating average of three bits e 2 3 FEC Code is a Hamming Code with 15 10 with hamming distance 4 e ARQ Automatic Repeat Request In some packets data field is retransmitted until a positive acknowledgment or timeout has reached Link Manager Protocol Link Manager sets up and configures the link Link Manager Protocol is the set of rules to discover and communicate with other devices There are several states of
93. s Hidden Master of Science Thesis Ali Osman ISIK 14 Communication in Networks Node Problem In contrast a station which senses only RTS can start a transmission As long as it does not collide with CTS it can finish its transmission successfully and this property solves Exposed Node Problem In spite of these improvements a collision can still occur when two stations send RTS at the same time to a mutual receiver CSMA CA is used in WLAN 802 11 Unlike the CSMA CD CSMA CA is also efficient under heavy traffic There are some other CSMA methods for wireless LANs which can slightly change from CSMA CA to improve the throughput Logical Link Control Layer The LLC sublayer acts as an interface between the MAC sublayer and the Network Layer This sublayer multiplexes the MAC protocols while transmitting and decoding while receiving It encapsulates the Network Layer data packets into the frames It also provides optional flow and error control in addition to the Transport Layer In wireless networks this control is used with retransmission of the erroneous packets however in LAN only error detection and canceling of the packet are provided In most of the network protocols IEEE 802 2 standard is used as the LLC protocol Now we will explain this standard in detail IEEE 802 2 This standard provides a common interface for upper layers to deal with any kind of MAC protocols It provides three types of operation for data communication e Typ
94. sage received from ALPHA1 lt SendToOtherRobot gt lt RECEIVER gt ALPHA4 lt RECEIVER gt lt FUNCTION gt lt SendToOtherRobot gt lt RECEIVER gt ALPHA5 lt RECEIVER gt lt FUNCTION gt lt TurnOffMonitor gt lt FUNCTION gt lt SendToOtherRobot gt lt FUNCTION gt lt SendToOtherRobot gt Message sent to ALPHA4 lt SendToOtherRobot gt lt RECEIVER gt ALPHA5 lt RECEIVER gt lt FUNCTION gt lt TurnOffMonitor gt lt FUNCTION gt lt SendToOtherRobot gt Output of ALPHA Message received from ALPHAS lt SendToOtherRobot gt lt RECEIVER gt ALPHA5 lt RECEIVER gt lt FUNCTION gt lt Turn0OffMonitor gt lt FUNCTION gt Ali Osman ISIK Master of Science Thesis 4 4 Handling Relayed Messages 67 lt SendToOtherRobot gt Message sent to ALPHAS lt Turn0OffMonitor gt Output of ALPHA5 Message received from ALPHA4 lt Turn0ff Monitor gt Monitor is turned OFF As we see from the outputs we used nested SendToOtherRobot functions and inside the inner one we send the real function In the output of ALPHA3 we see that message is parsed and function sub document is sent to the receiver In the output of ALPHA4 we see that message is parsed again and function sub document is sent to receiver ALPHA5 In the output of ALPHAJ3 message received parsed and TurnOffMonitor function that we added to listoffunctions h is called We also saw visually that the monitor of ALPHAS5 was turned off 4
95. sion detection is an analog process and the station s hardware should be able to listen the cable while transmitting Collision can be detected by comparing the power of received and transmitted signal Worst case collision detection time can be found as follows e Assume that at time 0 station A starts transmission and let the propagation time between two furthest stations is 7 e Assume that at time 7 e the furthest station B starts transmission unaware of the channel is busy e Then a collision occurs at time 7 and station B detects the collision aborts the trans mission and sends noise bursts to let every station know about the collision e At time 27 station A sees the noise burst and detects the collision So worst case collision detection time is two times the maximum propagation delay in the network This is the reason why there should be a minimum frame length in a network If a station tries to transmit a short frame less that 27 then the sender incorrectly concludes the transmission is successful before it detects the collision Hence if we notate the transmission time of a frame on the cable as t t should be larger than 27 for collision detection Note that 7 increases with the distance of nodes in the network and decreases with the signal speed t increases with the number of bits in a frame and decreases with transfer rate Generally the signal speed in the cable and transmission rate is fix and there is a trade of
96. source destination pair the graph is disconnected Connectivity is an important measure for the robustness of a MAS and it makes easier to study and understand the properties of the system A disconnected graph is union of two or more connected graphs These disjoint subgraphs are called connected components of graphs The largest of these components is called giant component When ratio of giant component size and the graph size is a measure for connectivity When it is closed to 1 the graph is closed to a connected graph and can be approximated to a connected graph to make the study easier In a graph hop count specifies the number of hops on the path between a source and a destination It is used to measure some concepts in graph theory such as Small World Property This is one of the most important phenomenon in connectivity of networks A network is said to have the small world property when the hop count in that network is not strongly affected by an increase in the network size Many networks in real life show small world property It also refers to the concept that everyone is connected to everyone else in the world by only six degrees of separation or six sets of acquaintances Clustering coefficient of a node is the ratio between the actual number of links between the neighbors of node 7 and the maximum possible number of links between these neighbors This coefficient is a measure to which nodes in a graph tend to cluster
97. ss from ALPHA1 to ALPHA2 Now we test the same SendX ML Message function with another message which intends to call a getter function of ALPHAJ First we prepared the XML message and sent it to ALPHAS The output including the XML message is as follows Sending message to ALPHA5 Message sent to ALPHA5 lt GetTime gt lt HANDLER gt lt FUNCTION gt PrintMessage lt FUNCTION gt lt HANDLER gt lt GetTime gt Message received from ALPHA5 lt PrintMessage gt lt STRING gt Localtime is 18 12 23 lt STRING gt Ali Osman ISIK Master of Science Thesis 4 1 Testing Basic Functions 61 lt PrintMessage gt Localtime is 18 12 23 If we look at the XML message in the output we see that the function tag is sent along with the handler information since Get Time is a getter function As we explained in the previous chapter getter functions are sending some data back and handler function is the function handling this data on the sender side In this particular example ALPHA1 calls Get Time function of ALPHAJ3 and it also tells how it will handle the time On ALPHA5 side GetTime function reads the clock of the PC ands prepares XML message to send back to ALPHA1 with the handler information As we see ALPHA1 receives message form ALPHAS which says to call its PrintMessage function As a result ALPHA1 asked for the time of ADPHA5 and told that it will print the time As we see on the last line ALPHA1 prints the
98. t Windows XP Home Edition x86 operating system The network is created with a wireless router and the PCs are connected to the network via their wireless adapters In order to build and package the DRL we used CMake open source build system CMake is a family of tools designed to build test and package software It is used to control the software compilation process using simple platform and compiler independent configuration files It also generates native makefiles to be used in the compiler environment of the choice 20 Since we want the DRL to run on both Microsoft Windows and Unix we used Cygwin to compile and test our library Cygwin is a Unix like environment and command line interface for Microsoft Windows Cygwin provides native integration of Windows based applications data and other system resources with applications software tools and data of the Unix like environment 21 The PCs are named as ALPHA 1 ALPHA2 ALPHA3 ALPHA and ALPHAJ These names are used to identify the robots within the network Random port numbers are assigned to each robot before running the DRL A user manual for compiling the library and running applications is given in Appendix A Four different test applications are implemented to test the capabilities of the DRL In Section 4 1 some commands are executed by one robot to test some of the basic functions of the library In Section 4 2 we implemented an application that synchronizes the clock of the robot by ex
99. t with server s IP and new port to start the communica tion as a client Then the socket is deleted to end the current session In Figure 3 6 we see simple connection request accept process in 7 steps After Robot sends the connection request Robot1 detects the connection request determines the IP of the requester and finds corresponding RobotComm Robot2Comm in this case Then Robot1 creates a new port and sends it to Robot2 Then Robot2 finds Robot1 Comm in its list After Robot2Comm of Robot1 creates the server Robot1Comm of Robot connects to the server and becomes the client In Figure 3 7 we see the flow chart of the process running on Robot1 side Server Client Communication Process Server Client communication process is at RobotComm level and takes care of the actual data transfer between agents Although we call this process as server client communication both server and client are able to send and receive data and there is no difference as communication partners After creation of sockets at both sides both sockets starts to listen for incoming data As we explained earlier the messages in this communication are also encoded with XML When a Robot wants to send a message to another Robot first communication is provided between Robots if there is not Then the corresponding RobotComm is found and this Robot Comm sends the message over its socket Since the corresponding RobotComm of the other Ali Osman ISIK Master of Scie
100. the task is complete and both RobotComm s keep on listening for further messages In Figure 3 9 we see flow chart of the process on Robot2Comm Robot1 side Robot2 wants to change location of Robot1 List of Robot1 s Functions Robot1 A Robot2 Y Y Y gt 4 Function is found and called 1 Send data to Robot1 Y A Y y Z 2 Send data y Robot2Comm i Robot1Comm y pr F 3 Receive and parse Jata 5 Agent1 s location is changed to X Y Agent1 Figure 3 8 Server Client Communication Process 3 2 Main Functions of Distributed Robotics Library Before explaining the functions some structures variables and important member variables of both of the classes of the DRL should be presented e message_handler This is a structure consists of function name pointer to that function and a pointer to an object The significance of third variable will be explained Master of Science Thesis Ali Osman ISIK 48 Implementation of the Distributed Robotics Library Robot2Comm Is there a matching function Call the function Figure 3 9 Flow Chart of Server Client Communication on Robotl Ali Osman ISIK Master of Science Thesis 3 2 Main Functions of Distributed Robotics Library 49 later A message handler is basically used to keep the pointer of a function of the robot e ADV_ PORT This is the advertisement port which is used by every robot to send and listen adver
101. time of ALPHA In figure 4 2 we explain the communication process better with the steps The rest of the handling functions such as saving received data are not implemented and left to the library user The whole output of ALPHAJ3 is as follows Al1i ALPHA5 console New Advertisement detected from ALPHA4 New Advertisement detected from ALPHA1 New Advertisement detected from ALPHA3 New Advertisement detected from ALPHA2 Connection Request detected from ALPHA1 Server started Message received from ALPHA1 lt GetTime gt lt HANDLER gt lt FUNCTION gt PrintMessage lt FUNCTION gt lt HANDLER gt lt GetTime gt Message sent to ALPHA1 lt PrintMessage gt lt STRING gt Localtime is 18 12 23 lt STRING gt lt PrintMessage gt The last tested function is the Stop Communication function of the library In the output list of robots is given after ALPHA 1 stops communication with ALPHA 2 Dropping connection with ALPHA2 List of the robots in the network ALPHA3 Not Connected Master of Science Thesis Ali Osman ISIK 62 Experimental Evaluation of Distributed Robotics Library 8 ALPHAS s Time is printed List of ALPHA1 s Functions List of ALPHAS s Functions MoveFirstLeg MoveFirstLeg GetTime GetTime 4 Time value is calculated 1 Send XML message Printessage PrintMessage to ALPHA5 5 Send XML message to ALPHA1 7 PrintMessage is called 3 GetTime is called 2 Sending XML messa
102. tion level where all the information is saved in huge databases they try to obtain more reliable DCSs Another model which is proposed by Aguilar and Cerrada 3 is a model based on a hierarchical architecture to be used in automating industrial processes In this hierarchy each level consists of a group of different tasks to be carried out within the control system and each level is using and generating different information Studies on distributed control greatly benefit from the utilization of Multi Agent Systems MASs since they provide a platform for simulation and testing of various hypotheses based Master of Science Thesis Ali Osman ISIK 2 Introduction on the principle of distributed artificial intelligence 4 Thus we can say that MASs are the new structure of decentralized control systems Distributed control MAS approaches have been used in various domains such as unmanned air terrestrial underwater vehicles search and rescue scenarios collective robotics social cognition etc 4 In some of the applications control can also be distributed over agents and humans For instance Koes and Nourbakhsh 5 have studied hybrid rescue group systems based on humans software agents and autonomous robots What they describe is a coordination architecture capable of quickly finding an optimal solution to the combined problems of task allocation scheduling and path planning subject to system constraints However Trianni and Dorigo
103. tisements The following variables are member variables of the class DistributedRoboticsLib e hostname The unique hostname of the Robot e port The port of the Robot which is used to contact it It does not have to be unique In our applications e info The information of the Robot which can consist of available functions or member data function_ list This is a list of type message_handler consisting of all the message handlers of the Robot e robot_comm_ list This is a list of type DistributedRobotCommLib consisting of the RobotComm s for the other Robots on the network The following variables are member variables of the class DistributedRobotCommLib e hostname The hostname of the Robot to be communicated with e ip_ address The IP address of the Robot to be communicated with e port The port of the Robot to be communicated with e parent_ robot A pointer to the Robot who owns this RobotComm e is connected A boolean variable which keeps track of the connection to the corre sponding Robot e watch_ dog An integer value decreasing by 1 every second It is used as a watchdog timer When it becomes 0 Robot removes RobotComm from the list and RobotComm is deleted 3 2 1 DistributedRoboticsLib Class Functions Advertise Advertise is a thread which starts running with the creation of the Robot it advertises the Robot to the network For the advertisement the hostname and the port of the Robot are encoded
104. together This leads us to the concept of Local Correlation Let node i be connected to node j If the probability of node 7 being connected to the neighbors of node j is higher than the probability of node 1 being connected to other nodes in the network we say that edges are locally correlated Local correlation is a factor that increases the clustering coefficient Different levels of connectivity can be utilized to analyze systems By definition k connectivity means the minimum clustering coefficient of a node is k In other words minimum number of links needed to be removed to get a disconnected graph is k In most of the times 1 connectivity is enough to analyze however if someone wants to set up more reliable networks upper connectivity levels can be considered Basic Graph Models For different MASs different graph models can be the best suited for graphical representation In this part we are looking for the best model for wireless networks Note that wireless Master of Science Thesis Ali Osman ISIK 8 Communication in Networks networks show small world property has local correlation and randomly distributed over a network The channels between nodes are RF channels and log distance path loss model is assumed In this model RF signal level decreases with the distance so the coverage of a node is limited In Erdos and Renyi random graph model We denote a random graph by G N where N is the number of nodes in the g
105. ue For instance assume there is a user wants to set up a local network with his camera printer and computer To assign an individual IP he has to set up Dynamic Host Configuration Protocol DHCP or do it manually To assign local names he has to Master of Science Thesis Ali Osman ISIK 34 Communication in Networks Wi Fi Bluetooth ZigBee Frequency 2 4GHz bands AAGE 868 915 MHz Stack size 1Mb 20kb 250kbps 2 4GHz Raw data rate 11 Mbps 40kbps 915MHz 20kbps 868MHz 16 2 4GHz 11 14 79 10 915MHz 1 868MHz Digital Digital Audio Key Value Pairs Number of channels Data types Digital Inter node tofdevices 32 8 255 65535 Medium hours Very low years requirements on one batte on one batte on one batte penetration Computer amp phone PEE Best Internet inside Low cost control applications buildings and monitoring peripherals Figure 2 9 Comparison of Different Wireless Technologies set up Domain Name System services DNS or do it manually Thus he is expected to be a network expert to set up a simple local network It is analogous to being expected to be a mechanic to drive a car To solve this problem a new software was implemented Zero Configuration Networking Bonjour as Apple calls it provides a three layer foundation to enable hardware and software makers to produce great products compatible to set up easy networking 16 The first layer is getting an IP a
106. undirected graphs as follows 8 e Two vertices u and v in a graph G are called adjacent or neighbors in G if u v is an edge of G e The topological structure of a graph is represented by the adjacency matrix Let N be the number of vertices Then the adjacency matrix will be N x N where each element Qij is either 1 if there is a connection or 0 if there is no edge connects vertices 7 and j e The degree of a vertex in a graph is the number of edges incident with it except that a loop at a vertex contributes two to the degree of that vertex The degree of the vertex v is denoted by deg v or d v The sum of degrees in a graph is double of the edges Ali Osman ISIK Master of Science Thesis 2 1 Modeling of Multi Agent Systems 7 e A vertex of degree zero is called an isolated vertex e A complete graph has an edge between every pair of vertices e A weighted graph has weight on its edges which can be analogous to the real distance between corresponding vertices Connectivity Many problems can be modeled with paths formed by traveling along the edges of graphs For instance the problem of determining whether a message can be sent between two computers using intermediate links can be studied with a graph model A path is a sequence of edges that begins at a source vertex and ends at a destination vertex in a graph A graph is connected if there is a path between every pair of vertices When there is no path for at least one
107. uni cation facilities simultaneously the TCP provides a set of addresses or ports within each host Concatenated with the network and host addresses from the internet communi cation layer this forms a socket A pair of sockets uniquely identifies each connection Logical Connections The reliability and flow control mechanisms described above require that TCP initializes and maintains certain status information for each data stream The combination of this status including sockets sequence numbers and win dow sizes is called a logical connection Each connection is uniquely identified by the pair of sockets used by the sending and receiving processes Full Duplex Each TCP connection supports a pair of byte streams one flowing in each direction TCP Header TCP data is encapsulated in an IP datagram The Figure shows the format of the TCP header Its normal size is 20 bytes unless options are present Each of the fields is discussed below The SrcPort and DstPort fields identify the source and destination ports respectively The sequence number identifies the byte in the stream of data from the sending TCP to the receiving TCP that the first byte of data in this segment represents The Acknowledgement number field contains the next sequence number that the sender of the acknowledgment expects to receive This is therefore the sequence number plus 1 of the last successfully received byte of data This field is valid only if the ACK fl
108. uter The main focus of this thesis is implementing a software infrastructure for communication of these netbooks in other words providing a software library to be used in a Distributed Robotics System DRS The library should be portable running in both Microsoft Windows and Unix operating systems By running parallel to the user applications the library should automatically discover other robots in the network to provide simple communication for data transfer to be used by these applications This kind of a communication library provides set of rules to create a common language for communication of different robots in a network 1 2 Thesis Outline In the first part of Chapter 2 we give background information about MASs their modeling and communication aspects The next part of Chapter 2 gives the background information about High Level Communication Protocols We start with the 7 layer Open Systems Interconnec tion OSI model which is the general model of a communication library then we explain ZigBee Wi Fi and Bluetooth communication protocols In Chapter 3 after introducing our Ali Osman ISIK Master of Science Thesis 1 2 Thesis Outline 3 Figure 1 1 Distributed Robotics Lab Distributed Robotics Library DRL we explain the main functionalities of the library Then by the help of UML sequence diagram we explain the execution of the library In Chapter 4 we give a demonstration of the library on a testbed by setting up a
109. ve s access code it goes into Master response state and returns a packet to synchronize the slave with itself Then the master goes to connected state e Once the slave receives the message it goes to connection mode too Then the master sends a poll message to check the connection and awaits a reply packet LMP_AU_RAND LMP_SRES With these messages authentication is done The master node sends LMP_ AU_ RAND message that contains a random number the challenge to the recipient After a specific algo rithm recipient sends LMP_SRES message and successful pairing of two devices is achieved Once the node is connected to the master it can be in the following states e Active mode is the mode that a slave is connected and listening the traffic e Hold mode is the mode where only synchronous links are active e Park mode is the mode where a slave cannot communicate with the master It is only active at certain slots to synchronize with the master and at the other slots it is sleeping e Sniff mode is for energy saving and it can affect ACL links since they are asynchronous Logical Link Control and Adaptation Protocol L2CAP This layer has three major functions e First it accepts 64 KB packets from the upper layers segment them for transmission and reassemble them for upper layers e It is responsible for multiplexing and demultiplexing of multiple packets i e determin ing which upper layer protocol to hand the packets after reassemb
110. which http identifies the protocol tas is a node in tudelft nl domain and tudelft is node in the nl domain When someone enters this URL DNS server is connected the name is resolved and searched hier archically and the node is found In Multicast DNS local is used in the namespace ali local to distinguish it from the global addresses In this case the name ali is sent to all nodes in the network and when a node sees a query for its own name it answers the query Instead of a DNS server additional software can handle this property Unlike the IP assignment the host name is set manually Once the local name is created a local Multicast DNS address record maps the names to the IP addresses To check the uniqueness again probing technique is used This time three queries are sent and if there is no conflict uniqueness is verified and the host announces its name to the network When there is a conflict a prompt message is sent and user is asked to change the host name Master of Science Thesis Ali Osman ISIK 36 Communication in Networks Service Discovery The last technology that Zeroconf uses is DNS Service Discovery DNS SD which lets you know the available services on the network without knowing the names or IP addresses of devices To use this service there should be link local Multicast DNS or global Unicast DNS should be available The service discovery software has two main responsibilities enumerating the list of services t
111. y useful in client server situations Since it gener ally assumes there is no error or correction is not necessary it is preferred in time sensitive applications where transmission delay is not acceptable The common applications of UDP are Domain Name System DNS streaming media applications such as Voice over IP VoIP Trivial File Transfer Protocol TFTP Transmission Control Protocol TCP In this subsection some texts are cited verbatim from the book 13 in order to maintain fidelity to the original referenced document TCP provides a connection oriented reliable byte stream service The term connection oriented means the two applications using TCP must establish a TCP connection with each other before they can exchange data TCP provides following facilities Ali Osman ISIK Master of Science Thesis 2 2 Modeling of Communication Libraries 19 Stream Data Transfer TCP transfers a contiguous stream of bytes by grouping the bytes in TCP segments and passing them to IP for transmission to the destination Reliability TCP assigns a sequence number to each byte transmitted and expects a positive acknowledgment ACK from the receiving TCP If the ACK is not received within a timeout interval the data is retransmitted The receiving TCP uses the sequence numbers to rearrange the segments when they arrive out of order and to eliminate duplicate segments Flow Control To allow for many processes within a single host to use TCP comm
112. yer is used by Bluetooth Direct Sequence Spread Spectrum DSSS is used by ZigBee DSSS based physical layer supplies high throughput and broad coverage while FHSS based physical layer is suitable for highly multipath environments In Wi Fi physical layer DSSS is used to achieve high throughput In this physical layer ISM band 2 4 2 4835 GHz is used and 14 channels of 22 MHz separated by 5 MHz are defined In early versions OFDM Orthogonal Frequency Division Multiplexing technique is used for modulation and 6 54 Mbps throughput could be achieved Typical radio range of Wi Fi is 100 meters and with 6 54 Mbps data rate Wi Fi is more suitable for indoor high rate applications For the LLC layer Wi Fi uses same properties that an LLC 802 2 layer and allows to connect a WLAN to any other LAN of the IEEE family MAC Layer Channel Access In WI FI superframe structure is similar to the one in ZigBee The superframe starts with a beacon then contention free period CFP and contention period CP follows In CFP point coordination function PCF is selected for transmission which is similar to GTSs in ZigBee This function is conceived for the data transmission with time constraints such as the real time traffic voice video Basic access method in WI FI is applied during the CP and is called distributed coordination function DCF CSMA CA protocol is used as an access method for DCF To solve hidden and exposed node problems it is supplemented by
113. yperText Transfer Protocol HTTP File Transfer Protocol FTP Domain Name System Protocol DNS etc Master of Science Thesis Ali Osman ISIK 24 Communication in Networks 2 2 2 Interface OSI Reference Model does not specify any programming interfaces other than abstract service specifications The protocol specifications given in the OSI model define the interfaces between communicating agents however there can be implementation specific software interfaces in each agent For instance Microsoft Windows Winsock and Unix s Berkeley sockets and System V Transport Layer Interface are interfaces between Layer 5 and Layer 4 2 3 High level Communication Protocols 2 3 1 Bluetooth Protocol In 1994 Erickson Mobile Communications became interested in low cost low consumption radio interface between mobile phones and other devices In 1998 IBM Intel Nokia and Toshiba joined the project They formed a SIG Special Interest Group to develop a short range low power low cost wireless radios for communication devices and the project was named Bluetooth The first specifications was published in 1999 by Bluetooth SIG and it was the basis of IEEE 802 15 standards which is about Wireless Personal Area Networks WPAN Architecture Communication in bluetooth is based on master slave principle A master and up to seven slaves within a distance of 10 meters form a cell called piconet Several piconets can overlap and form a scatternet

Download Pdf Manuals

Related Search

Ali Master Thesis

Ali-Master-Thesis

Contents

Download Pdf Manuals

Related Search

Related Contents