DeCoLibrary User Guide
Contents
1. This code first defines a string array with default settings as shown below e gui false disables the JCC e welcome false disables printing of the welcome message with the platform setup e cli false disables the command line interface e printpass false disables printing of platform password After that a new array is created First one is filled with the default arguments and then second one is filled with the args which are supplied to the main method Therefore in case of conflicts the arguments supplied from the outside override the default settings Make sure that in the createPlatform method the newargs array is passed instead of the args Otherwise the default settings will be ignored package deco jadex examples import jadex base Starter import jadex bridge IExternalAccess import jadex bridge service RequiredServicelInfo import jadex bridge service search SServiceProvider import jadex bridge service types cms IComponentManagementService import jadex commons future IFuture import jadex commons future ThreadSuspendable public class ExecuteMain public static void main String args String defargs new String oui false welcome false pli false eprintpass false String newargs new String defargs lengthtargs length System arraycopy defargs 0 newargs 0 defargs length System arraycopy args 0 newargs defargs length args length2. IFuture lt IExternalAccess gt platfut Starter createPlatform newargs ThreadSuspendable sus new ThreadSuspendable IExternalAccess platform platfut get sus System out printin Started platform platform getComponentIdentifier IComponentManagementService cms SServiceProvider getService platform getServiceProvider IComponentManagementService class RequiredServiceInfo SCOPE PLATFORM get sus cms createComponent deco jadex examples HelloDeCo class null get sus Figure 8 Main Class for DeCoAgent Execution In order to search component management service and to create an access point for the service mechanism we use the static helper class SServiceProvider from package jadex bridge service search The result of the getService method is a future of the corresponding service type The application is still running on the Java main thread therefore it is safe to use the thread suspendable again for blocking until the search result is available We pass our DeCoAgent class name with package to the createComponent function of the component management service Finally we run ExecuteMain class as a java application and our HelloDeCo agent is started to run on platform Jadex The execution s output of the agent will be shown in Eclipse console panel as Hello World 3 DeCoAgent Design and Implementation DeCoA gent implementation needs a deliberative coherence m
3. Java Application o Select type any character any String TZ TimeZone v s Matching items Q scollection jadex commons collection Q sutil jadex commons Q Schema com sun java help search Q ScriptDemo org codehaus janino samples ScriptEvaluator org codehaus janino Q ShippingCost org codehaus janino samples Q simpleCompiler org codehaus janino Q startScenario jadex micro testcases semiautomatic remoteser starter jadex base Q Statistic cern colt matrix doublealgo Q StatusBar jadex base gui Q streamDecompressor com sun java help search Q stringTemplateTreeView org antir stringtemplate misc Q synchronizationTimer EDU oswego cs dl util concurrent misc m ha jadex base Figure 3 Select Starter Class for Starting Jadex Platform Hit Ok to start the Jadex platform The next time you want to start the platform you do not have to repeat the above steps Just choose the Starter entry from the run history which eclipse generates automatically If you successfully start the Jadex platform the Jadex control center JCC will appear in the screen as seen in Fig 4 The JCC is a management and debugging interface for the Jadex platform and components that run on it To execute any application you need to add the corresponding path to the JCC project We will now set up the platform for starting some examples Right click in the upper le
4. in PLans annotation Plan trigger Trigger goals CleanerDeCo KeepBatteryLoaded class body Body KeepBatteryLoadedPlan class Plan trigger Trigger goals CleanerDeCo ReturnHome class body Body ReturnHomePlan class Figure 44 Plan Headings Now run the ExecutionMain java file as a java application and observe the results
5. our agent definition file CleanerDeCo java Goal excludemode ExcludeMode Never succeedonpassed false relations Relations facilitations Facilitate units Move class degree 1 priority 1 0 public class MoveAround public MoveAround Figure 16 MoveAround Goal Definition Goal annotation fields and constants except relations and priority fields are about BDIA gent and its documents can be found at http www activecomponents org Now we must implement a plan that can be executed when MoveAround goal is adopted Create a new java document and name it as MoveAroundPlan java We want our agent to pick a random position and move there iteratively So our plan will create a random position and create a new Move goal to this target position In a plan body we can dispatch sub goals for execution by using plan API In Fig 17 we define rplan as our plan API and we dispatch Move sub goal using it This dispatching means that Move goal which is a subgoal of the MoveAround goal will be executed Add the following MoveAroundPlan definition to our MoveAroundPlan java file package deco jadex examples cleaner import deco jadex bdiv3 annotation Plan import deco jadex bdiv3 annotation PlanAPI import deco jadex bdiv3 annotation PlanBody import deco jadex bdiv3 annotation PlanCapability import deco jadex bdiv3 runtime IPlan import deco jadex examples cleaner CleanerDeCo Move import jad
6. relationships ReLation annotation had facilitations and incompatible fields priority field defines the inherited priority of the current goal and takes double parameter between 0 0 1 0 To represent facilitation in DeCo model an excitatory link is created between goals proportional to the indicated degree facilitations field take one or more Facilitate annotation object for defining facilitation relationships between goals It have two fields units and degree We define goals in units field that facilitate current goal units field take goal classes as a parameter and can take one or more parameter degree represents the degree of the facilitation degree field takes double parameter between 0 0 1 0 To represent incompatibility in DeCo model an inhibitory link is created between goals for the indicated degree incompatible field takes one or more IncompatibLe annotation object for defining incompatible relationships between goals It has two fields unit and degree We define goal in unit field that is incompatible with current goal unit field takes goal classes as a parameter and can take only one goal class as a parameter degree represents the degree of the incompatibility degree field takes double parameter between 0 0 1 0 You can void Goal annotation fields Then no relationship will be defined for current goal and i
7. 0 rc7 E Use default location Location C Files Checkouts jadex target jadex 2 0 rc7 JRE Use an execution environment JRE JavaSE 1 6 Use a project specific JRE jdk1 6 0_25 Use default JRE currently jdk1 6 0_25 Configure JREs Project layout Use project folder as root for sources and class files Create separate folders for sources and dass files Configure default Working sets E Add project to working sets Select The wizard will automatically configure the JRE and the project layout based on the existing source Figure 1 Creation of Java Project in Eclipse Click Finish then the project will be created Your project should now appear IN your package explorer This project will be used as a basis for your own development projects To make the Jadex libraries accessible to other projects it is necessary in Eclipse to export them Right click on project choose Build Path gt Configure Build Path Go to the Order and Export tab choose Select All and hit Ok ri Java Build Path Resource Builders Source Projects Libraries Order and Export Java Build Path Build class path order and exported entries Java Code Style Exported entries are contributed to dependent projects Java Compiler Java Editor Javadoc Location Project Facets Project References Refactoring History Run Debug Settings Task Repository Task Tags Validation Wi
8. 2 jadex Javadoc Location jadex 2 4 Top gt Papyrus e activation 1 1 jar jadex 2 4 lib Project References amp antlr 2 7 7 jar jadex 2 4 lib Bottom i Refactoring History amp antlr runtime 3 1 3 jar jadex 2 4 lib Run Debug Settings fs asm4 4 0 jar jadex 2 4 lib Select All Task Tags heprov elk saja POSEN Sey XML Syntax amp s bridj 0 6 2 jar jadex 2 4 lib Sa amp collections generic 4 01 jar jadex 2 4 lib G8 colt 1 2 0 jar jadex 2 4 lib amp concurrent 1 3 4 jar jadex 2 4 lib de huxhorn sulky 3rdparty jlayer 1 0 jar jadex 2 4 lit Figure 6 Change the Order of the DeCoAgentLib The configuration of our environment is completed Now we can start to implement our first DeCoA gent A 2 Implementing a DeCoAgent 1 First DeCo Agent Choose DeCoExamples project from package explorer and create a new package called deco jadex examples DeCoAgent is a BDIv3 agent and most of its properties belong to Jadex BDIAgent For BDIAgent detailed documentation you can visit Jadex Active Components site http www activecomponents org bin view BDI V3 Tutorial 01 Introduction The agent definition file is a normal Java class that uses Agent annotation to state that is an agent It is required that the Java file ends with DeCo Otherwise it will not be recognized as DeCoAgent Create a new empty Java class and name it as HelloDeCo class Add Agent annotation before HelloDeCo class d
9. DeCo MoveAround class body Body MoveAroundPlan class public class CleanerDeCo Figure 18 Goal and Plan Binding When a BDIAgent starts the execution it needs a top level goal In our agent definition file add a body function annotated with AgentBody and dispatch MoveAround goal as a top level goal Our deliberative coherence model sets activations of the units according to desirability of the model and we set our DeCoModel desirability as 0 05 AgentBody public void body agent setDesirability 0 05 agent dispatchTopLevelGoal new MoveAround Figure 19 AgentBody For execution of our agent reopen the ExecuteMain class file and change the create component line as shown in Fig 20 cms createComponent deco jadex examples cleaner CleanerDeCo class null get sus Figure 20 Running the CleanerDeCo Run ExecuteMain class file as a java application and observe the console output c Case Study 2 1 Scenario Description According to the second case study cleaner will have dirtiness sensation in addition to the first case study It will move around randomly and sense the dirtiness from the environment 2 DeCo Model Design We currently have a cleaner agent that can move around Now we want our agent to look for waste Firstly our cleaner needs an action Sense to sense the dirtiness and move around to look for waste We will add our model LookForWaste goal and this goal wi
10. DeCoAgent Deliberative Coherence Driven Agent User Guide DeCoAgentLibrary v1 0 30 June 2014 This document presents a user manual for DeCoAgent Library package Any person who want to create a DeCoAgent can use this tutorial In the first part of the user manual environment configuration of the DeCoAgent Library will be presented In the second part a Cleaner World agent a famous example in intelligent agent literature design and implementation with the usage of DeCoAgent Library will be presented A 1 Development and Execution Environment Configuration 1 Prerequisites e Download and install a recent Java environment from http java sun com javase downloads if not installed e Download and install a suitable Eclipse distribution gt 3 5 from http www eclipse org downloads if not installed e Download Jadex 2 3 distribution zip from http jadex agents informatik uni hamburg de xwiki bin view Download Available Packages and unpack it to a place of your choice e Download DeCoAgent package DeCoAgentLib jar from https sourceforge net projects decoagentlibrar 2 Eclipse Environment Setup Start Eclipse IDE and start the New Java Project wizard disable the Use default location checkbox and browse to the directory where you unpacked the Jadex distribution 6 New Java Project Create a Java Project Create a Java project in the workspace or in an external location Project name jadex 2
11. ass degree 1 public class ReturnHome public ReturnHome Figure 39 ReturnHome Goal Definition In order to define our plan for ReturnHome goal create a new java document named as ReturnHomePlan java In our ReturnHomePlan body we will dispatch a move goal with our charging station position package deco jadex examples cleaner import jadex commons future ExceptionDelegationResultListener import jadex commons future Future import jadex commons future IFuture import deco jadex bdiv3 annotation Plan import deco jadex bdiv3 annotation PlanAPI import deco jadex bdiv3 annotation PlanBody import deco jadex bdiv3 annotation PlanCapability import deco jadex bdiv3 runtime IPlan import deco jadex examples cleaner CleanerDeCo Move Plan public class ReturnHomePlan PlanCapability protected CleanerDeCo capa PlanAPI protected IPlan rplan public ReturnHomePlan PlanBody public IFuture lt Void gt body final Future lt Void gt ret new Future lt Void gt double xDest 0 double yDest 0 IFuture lt Move gt fut rplan dispatchSubgoal capa new Move xDest yDest fut addResultListener new ExceptionDelegationResultListener lt CleanerDeCo Move Void gt ret public void customResultAvailable Move amt ret setResult null p return ret Figure 40 ReturnHomePlan java Definition Open the CleanerDeCo java file a
12. customResultAvailable MoveAround amt ret setResult null J return ret Figure 27 LookForWastePlan java Definition Add the code in the Fig 27 to the LookForWastePlan java file Until now we have defined our plans and goals Now open the CleanerDeCo java file and add the following plan headings in P Lans annotation Plan trigger Trigger goals CleanerDeCo LookForWaste class body Body LookForWastePlan class Plan trigger Trigger goals CleanerDeCo Sense class body Body SensePlan class Figure 28 Plan Headings Now run the ExecutionMain java file as a java application and observe the results d Case Study 3 1 Scenario Description According to the third case study cleaner will have vacuum actuator in addition to the second case study and will satisfy cleaning performance criteria It will move around randomly sense the dirtiness from the environment and vacuum the dirtiness 2 DeCo Model Design Until now our cleaner agent looked for waste by only moving around and sensing the environment Now we will add cleaning ability to our cleaner agent Our cleaning ability will be facilitated by vacuum action and sense actions Since the cleaner agent will not be able to handle both of LookForWaste and Clean goals we add incompatible relationship between these goals The appearance of the DeCo model activations when LookForWaste priority is 1 Clean priority is O left and LookForWaste priorit
13. d setter functions of these beliefs from menu Source gt Generate Getters and Setters Imported DeCoAgent libraries are sensitive and may be interfered with Jadex BDIv3 libraries Keep in mind that when you are implementing a DeCoAgent you should not import any BDIv3 library package deco jadex examples cleaner import deco jadex bdiv3 DeCoAgent import deco jadex bdiv3 annotation Belief import jadex micro annotation Agent Agent public class CleanerDeCo Agent DeCoAgent agent Belief private double xPos Belief private double yPos public double getxPos return xPos public void setxPos double xPos this xPos xPos public double getyPos return yPos public void setyPos double yPos this yPos yPos Figure 10 CleanerDeCo Agent Definition For now our agent have no goals We will add our first goal Move to our agent As we mentioned earlier DeCoAgent has all of the Jadex BDIV3 abilities You can define goals with Inner Classes or Classes We will define our goals with inner classes Create a new Inner Class and name it as Move Goals create basis of the deliberative coherence model creation In Jadex in order to define a goal you must use Goal annotation In DeCoAgent library Goal annotation have relations and priority fields in addition to the JadexBDIV3 abilities relations field take an ReLation annotation for defining incompatible and facilitation
14. e dirtiness b Case Study 1 1 Scenario Description According to the first case study cleaner will have no performance criteria a steering actuator and no sensor It will only move around randomly 2 DeCoModel Design Firstly we will create an agent that is only moving around Our goal will be MoveAround and our action will be Move MoveAround will only create a random position and post it to Move action Our deliberative coherence model will be as in Fig 9 When MoveAround goal priority is 1 both of the units will be activated left side model and when MoveAround goal priority is 0 both of the units will be deactivated right side model Desirability 05 Desirability 05 Mg M Figure 9 Exercise 1 Cleaner DeCo Model According to our created model when MoveAround priority is 1 MoveAround goal activation will be 0 54 and Move action activation will be 0 29 When MoveAround priority is 0 MoveAround goal activation will be 0 54 and Move action activation will be 0 29 3 Implementation Create a new package and name it as deco jadex examples cleaner Right click on the package create new java class file and name it as CleanerDeCo java Add Agent annotation before the class definition Add DeCoAgent API as an attribute for the agent Our agent currently will only move around We need two beliefs for keeping agent coordinates xPos yPos We add our beliefs to agent by BeLief annotation We generate getter an
15. e in synchronized Add the dirtiness as a double grid array and initialize the dirtiness of the room at the constructor Add a synchronized vacuum function to the class in order to provide the controlled access Add the following code Room java class file package deco jadex examples cleaner import java util Random public class Room protected static Room instance protected double dirtiness protected int size 100 public Room Initialize the room dirtiness dirtiness new double 100 100 for int i 0 i lt size i Random generator new Random int System currentTimeMillis for int j 0 j lt size j double dirt 1 generator nextGaussian 2 if dirt lt 0 dirt 0 dirtiness i j dirt public static synchronized Room getInstance if instance nul1l instance new Room return instance public double getDirtiness double x double y int xIdx int x size int yIdx int y size return dirtiness xIdx yldx public synchronized void vacuum double x double y double value int xIdx int x size int yIdx int y size dirtiness xIdx yIdx value Figure 25 Room java Definition Open the CleanerDeCo java file and add the LookForWaste goal as an inner class Our LookForWaste goal must be facilitated by the MoveAround and Sense goals Its priority will be 1 Add the following LookForWaste definition to the a
16. efinition This annotation states that this class is an agent Add a variable called agent to the agent class and add annotation Agent above the agent variable The field should be type of DeCoAgent This will let the engine automatically inject the DeCo Agent API to the plain old java object pojo agent class Add an agent body method that is automatically invoked when the agent is started Add a println line inside the agent body for console output package deco jadex examples import deco jadex bdiv3 DeCoAgent import jadex micro annotation Agent import jadex micro annotation AgentBody Agent public class HelloDeCo Agent DeCoAgent agent AgentBody public void body System out printin Hello World Figure 7 First DeCoAgent Example This first agent is not different from a classic BDIAgent and does not contain any DeCoAgent ability We implement this agent for testing our environment 2 Executing DeCoAgent Jadex Control Center only executes its inner assigned agent types MicroAgent BDIAgent DeCoAgent is an extension package for Jadex platform and cannot be executed within JCC But Jadex is a service based library and we can use its services with Jadex libraries We will execute our agents with a Java application that contains main function Create a new java class file named ExecuteMain class Add a main method to the class As explained above place the following code inside the method body
17. ex commons future ExceptionDelegationResultListener import jadex commons future Future import jadex commons future IFuture Plan public class MoveAroundPlan PlanCapability protected CleanerDeCo capa PlanAPI protected IPlan rplan public MoveAroundPlan PlanBody public IFuture lt Void gt body System out printin Cleaner started to execute Move Around plan final Future lt Void gt ret new Future lt Void gt double xDest Math random double yDest Math random IFuture lt Move gt fut rplan dispatchSubgoal capa new Move xDest yDest fut addResultListener new ExceptionDelegationResultListener lt CleanerDeCo Move Void gt ret public void customResultAvailable Move amt ret setResult null e return ret Figure 17 MoveAroundPlan Definition We implemented all the goals and plans but before running our agent we must bind goals and plans in agent definition file In CleanerDeCo java file after Agent annotation add plans in a PLans annotation PLans annotation takes one or more PLan annotation as an element A P Lan annotation have two fields First field is the goal definition class trigger object of the plan Second field is the plan definition class body of the plan Agent Plans Plan trigger Trigger goals CleanerDeCo Move class body Body MovePlan class Plan trigger Trigger goals Cleaner
18. ft area called the model explorer as it is used to browse for models of e g processes and choose Add Path S Jadex Control Center 2 0 rc7 28 July 2011 joc Alex PC_d59 File Help aa QGI3BS 253 RES AGAUMKORAB adex Settings ee Add Path Filename File Filter gt Configuration RA Add a new directory path package root to the project structure it C Auto generate Parent gt lt Flags O Start suspended C Master _ Daemon C Auto Shutdown start Reload Reset q B Alex PC_d59 o Z Services o kernels o rms amp awa o Lars PC_3f6 Lars PC_3f6 Figure 4 Add Path in JCC For executing a sample agent open the ib directory and select the file jadex applications micro jar You can now unfold the contents of the jar file and browse to the helloworld example in the jadex examples package After you select the HelloWorldAgent class in the tree you can start the process by clicking Start The component will be executed and it will print some messages to the eclipse console 3 Create and Setup a DeCo Agent Project Create a new Java project from File gt New gt New Java Project and name it as DeCoExamples Choose your project location or use default one Click Finish then your empty java project will be created and shown in package explorer New Java Pr
19. ge the priorities of the goals according to our created deliberative coherence model According to our agent model our agent should start to execute the Clean goal after sensing the dirtiness In order to implement this open the SensePlan java file and update the body of the plan as in Fig 36 PlanBody public IFuture lt Void gt body final Future lt Void gt ret new Future lt Void gt Room room Room getInstance double dirtiness room getDirtiness capa getxPos capa getyPos double batteryLevel capa getBatteryLevel System out printin Dirtiness dirtiness BateryLevel batteryLevel capa setSensedDirt dirtiness if dirtiness gt 1 capa agent setPriority Clean class 1 0 capa agent setPriority LookForWaste class 0 0 return ret Figure 36 Updated SensePlan Body Now run the ExecutionMain java file as a java application and observe the results e Case Study 4 1 Scenario Description According to the fourth case study cleaner will satisfy keep battery loaded performance criteria in addition to the third case study It will move around randomly sense the dirtiness from the environment vacuum the dirtiness and keep battery loaded 2 DeCo Model Design Until now our cleaner looks for waste in an environment and cleans the dirtiness From now on to make our agent a real cleaner we will add the charging ability to the cleaner We will add KeepBatteryLoaded goal to keep the bat
20. gent definition file Goal excludemode ExcludeMode Never succeedonpassed false relations Relations facilitations Facilitate units MoveAround class degree 1 Facilitate units Sense class degree 1 priority 1 0 public class LookForWaste public LookForWaste Figure 26 LookForWaste Goal Definition In order to create LookForWaste plan create a new java document named as LookForWasteP lan java package deco jadex examples cleaner import jadex commons future ExceptionDelegationResultListener import jadex commons future Future import jadex commons future IFuture import deco jadex bdiv3 annotation Plan import deco jadex bdiv3 annotation PlanAPI import deco jadex bdiv3 annotation PlanBody import deco jadex bdiv3 annotation PlanCapability import deco jadex bdiv3 runtime IPlan import deco jadex examples cleaner CleanerDeCo MoveAround Plan public class LookForWastePlan PlanCapability protected CleanerDeCo capa PlanAPI protected IPlan rplan public LookForWastePlan PlanBody public IFuture lt Void gt body System out printin Cleaner started to execute Look For Waste plan final Future lt Void gt ret new Future lt Void gt IFuture lt MoveAround gt fut rplan dispatchSubgoal capa new MoveAround fut addResultListener new ExceptionDelegationResultListener lt CleanerDeCo MoveAround Void gt ret public void
21. kiText wo activation 1 1 jar jadex 2 0 rc8 lib amp antir 2 7 7 jar jadex 2 0 rc8 lib sg antir runtime 3 1 3 jar jadex 2 0 rc8 lib Gs collections generic 4 01 jar jadex 2 0 rc8 lib ag colt 1 2 0 jar jadex 2 0 rc8 lib ic concurrent 1 3 4 jar jadex 2 0 rc8 lib cy de huxhorn sulky 3rdparty jlayer 1 0 jar jadex 2 9 gluegen rt 1 jar jadex 2 0 rc8 lib fag jadex applib bdi 2 0 rc8 jar jadex 2 0 rc8 lib icy jadex applications bdi 2 0 rc8 jar jadex 2 0 rc8 li ag jadex applications bdibpmn 2 0 rc8 jar jadex 2 0 xg jadex applications bpmn 2 0 rc8 jar jadex 2 0 rct cy jadex applications gpmn 2 0 rc8 jar jadex 2 0 rct a jadex applications micro 2 0 rc8 jar jadex 2 0 rc ng jadex bridge 2 0 rc8 jar jadex 2 0 rc8 lib na jadex commons 2 0 rc8 jar jadex 2 0 rc8 lib jadex javaparser 2 0 rc8 jar jadex 2 0 rc8 lib bs jadex kernel application 2 0 rc8 jar jadex 2 0 rc amp m TARERE RRRARRR RARATAN Figure 2 Export jars from Build Path To start the Jadex platform right click on the Jadex project which is in the package explorer Then choose Run As gt Java Application from the pop up menu Eclipse will search for main types startable main classes Select Starter from the package jadex base in the dialog Select
22. ll be facilitated by MoveAroundGoal and Sense action When we add the LookForwaste goal and Sense action to our model the activations of nodes will be as shown in Fig 21 When the priority of the LookForWaste goal is 1 all of the units become activated When the priority of the LookForWaste goal is 0 all of the units become deactivated So our agent has the capability of both moving around and sensing the dirtiness Desirability 05 Desirability 05 Figure 21 The Appearance of the DeCo Model Activations 3 Implementation We have created our DeCo model now we will implement our model in our agent definition file Open the agent definition file CleanerDeCo java We need two beliefs which are the sense and the environment where one of them is related with the amount of the dirtiness and the other is related with the time slot of the sense Add the sensedDirt time belief and their setter getter functions to our agent definition Belief private double sensedDirt Belief updaterate 200 private long time System currentTimeMillis public double getSensedDirt return sensedDirt public void setSensedDirt double sensedDirt this sensedDirt sensedDirt Figure 22 Beliefs to Sense the Environment We have added the necessary beliefs to our agent Now we will add the Sense goal Sense goal will be created for every time event To make this happen we use GoalCreatio
23. me it as KeepBatteryLoadedPlan java In our plan body we will dispatch a ReturnHome goal and wait for the agent to return to the charging station After agent returns to the charging station agent recharges itself When recharging is completed agent will change the priorities of the goals in order to start LookForWaste goal package deco jadex examples cleaner import jadex commons future Future import jadex commons future IFuture import deco jadex bdiv3 annotation Plan import deco jadex bdiv3 annotation PlanAPI import deco jadex bdiv3 annotation PlanBody import deco jadex bdiv3 annotation PlanCapability import deco jadex bdiv3 runtime IPlan import deco jadex examples cleaner CleanerDeCo Clean import deco jadex examples cleaner CleanerDeCo KeepBatteryLoaded import deco jadex examples cleaner CleanerDeCo LookForWaste Plan public class KeepBatteryLoadedPlan PlanCapability protected CleanerDeCo capa PlanAPI protected IPlan rplan public KeepBatteryLoadedPlan PlanBody public IFuture lt Void gt body System out printin Battery is low recharging is needed System out printin Returning to battery charge station final Future lt Void gt ret new Future lt Void gt rplan dispatchSubgoal capa new ReturnHome get double batteryLevel capa getBatteryLevel while batteryLevel lt 1 batteryLevel Math min batteryLevel 0 01 1 0 capa
24. nCondition annotation Add the following code to the agent definition file Goal public class Sense GoalCreationCondition events time public Sense id Figure 23 Sense Goal Definition To create Sense plan create a new java document named as SensePlan java In every execution of the SensePlan agent will take the environment and will sense the dirtiness of the environment s current position Add the following code to the SensePlan java package deco jadex examples cleaner import jadex commons future Future import jadex commons future IFuture import deco jadex bdiv3 annotation Plan import deco jadex bdiv3 annotation PlanAPI import deco jadex bdiv3 annotation PlanBody import deco jadex bdiv3 annotation PlanCapability import deco jadex bdiv3 runtime IPlan Plan public class SensePlan PlanCapability protected CleanerDeCo capa PlanAPI protected IPlan rplan public SensePlan PlanBody public IFuture lt Void gt body final Future lt Void gt ret new Future lt Void gt Room room Room getInstance double dirtiness room getDirtiness capa getxPos capa getyPos System out println Dirtiness dirtiness capa setSensedDirt dirtiness return ret Figure 24 SensePlan java Definition We need an instance of environment to store the dirtiness Create Room java class file and add a static instance of the room to get the room class anywher
25. nd add the KeepBatteryLoaded goal as an inner class Since the KeepBatteryLoaded goal is facilitated by ReturnHome goal we will define a facilitation relationship between them Additionally we will define 2 incompatible relationships between KeepBatteryLoaded LookForWaste and Clean goals Our KeepBatteryLoaded goal will be unique and created with recharge event Goal unique true deliberation Deliberation inhibits KeepBatteryLoaded class velations Relations facilitations Facilitate units ReturnHome class degree 1 incompatibles Incompatible unit LookForWaste class degree 1 Incompatible unit Clean class degree 1 priority 0 0 public class KeepBatteryLoaded GoalCreationCondition events recharge public KeepBatteryLoaded a public int hashCode final int prime 31 int result 1 result prime result getOuterType hashCode result prime result 0 return result public boolean equals Object obj boolean ret false if obj instanceof KeepBatteryLoaded KeepBatteryLoaded other KeepBatteryLoaded obj ret getOuterType equals other getOuterType return ret private CleanerDeCo getOuterType return CleanerDeCo this Figure 41 KeepBatteryLoaded Goal Definition After adding our KeepBatteryLoaded goal to our agent definition file Create a new java document and na
26. odel Before implementing our DeCoAgent we will describe the scenario of the cleaner world After the definition of goals and actions we will design our deliberative coherence model and finally we will implement our DeCoAgent In this document we will explain DeCoAgent capabilities within four exercises in expanding manner In each step we will add additional abilities to our agent a Cleaner World General Scenario Description A cleaner is an autonomous robot that can randomly move around can sense the dirtiness can clean the dirtiness and can recharge itself The cleaner agent description according to PEAS criteria presented in Table 17 The environment of the agent will be a room that have no object Table 1 PEAS Description of Cleaner World Clean Cleaner tries to keep its environment clean Performance When battery level below than a defined Keep Battery Loaded value it should return charging station to keep battery loaded Cleaner can only sense current position of the Partially observable environment where it possess Environment is not changing in Cleaner Static l reasoning process Environment The next state of the environment cannot be Stochastic determined with current state Steering is a continues state and continues Continues time problem Steering For moving around Actuators Spe Vacuum For cleaning the dirtiness Sensors Dirtiness Sensor For sensing th
27. oject Create a Java Project Create a Java project in the workspace or in an external location Project name Cleanerworld J Use default location Location home comp workspace Browse working sets T Add project to working sets Next gt conce A Figure 5 Create DeCoExamples Java Project DeCo Agent library needs Jadex libraries for execution Put DeCoAgentLib jar file under the Jadex project source library file For configuring our DeCoExamples project right click on project and choose Build Path gt Configure Build Path Go to the Libraries tab choose Add JARs and choose Jadex project from appearing window and add all library files to the project Then choose Add External JARs browse the DeCoAgent lib file that needs to be added the Jadex lib source and choose OK After importing the necessary libraries go to the Order and Export tab and change the order of the DeCoAgentLib jar for it to be above Jadex libraries as shown in Fig 6 Properties for Cleaner Java Build Path gt Resource gt ATL source Projects BALibraries eOrder and Export Builders Build class path order and exported entries Java Build Path Exported entries are contributed to dependent projects gt Java Code Style amp cleaner sre Up gt Java Compiler JRE System Library JavaSE 1 7 gt Java Editor amp DecoAgentLib jar home comp workspace
28. ort jadex commons future IFuture import deco jadex bdiv3 annotation Plan import deco jadex bdiv3 annotation PlanAPI import deco jadex bdiv3 annotation PlanBody import deco jadex bdiv3 annotation PlanCapability import deco jadex bdiv3 runtime IPlan import deco jadex examples cleaner CleanerDeCo Vacuum Plan public class CleanPlan PlanCapability protected CleanerDeCo capa PlanAPI protected IPlan rplan public CleanPlan PlanBody public IFuture lt Void gt body System out printin Cleaner started to execute Clean final Future lt Void gt ret new Future lt Void gt IFuture lt Vacuum gt fut rplan dispatchSubgoal capa new Vacuum 7 fut addResultListener new ExceptionDelegationResultListener lt CleanerDeCo Vacuum Void gt ret public void customResultAvailable Vacuum amt ret setResult null p return ret Figure 34 CleanPlan java Definition Until now we have defined our plans and goals Now open the CleanerDeCo java file and add the following plan headings in P Lans annotation Plan trigger Trigger goals CleanerDeCo Clean class body Body CleanPlan class Plan trigger Trigger goals CleanerDeCo Vacuum class body Body VacuumPlan class Figure 35 Plan Headings Changing the priorities of the goals is the key point of the DeCoAgent goal deliberation process When a scheduled action has been taken agent must chan
29. rt deco jadex examples cleaner Room import deco jadex examples cleaner CleanerDeCo Clean import deco jadex examples cleaner CleanerDeCo LookForWaste Plan public class VacuumPlan PlanCapability protected CleanerDeCo capa PlanAPI protected IPlan rplan public VacuumPlan PlanBody public IFuture lt Void gt body final Future lt Void gt ret new Future lt Void gt Room room Room getiInstance while capa getSensedDirt gt 0 1 room vacuum capa getxPos capa getyPos 0 025 rplan waitFor 100 get capa agent setPriority Clean class 0 0 capa agent setPriority LookForWaste class 1 0 return ret Figure 32 VacuumPlan java Definition We can reconfigure the agent DeCoModel from any plan by updating the priorities and relationships of the goals In order to access these functions we must use API of the DeCoA gent setPriority Class lt gt goal double priority function takes two parameters First one is for goal class definition and the second one is for priority Open CleanerDeCo java file Create a new goal as an inner class and name it as Clean Define our DeCoModel relations in relations field of the GoaL annotation Our Clean goal must be facilitated by Vacuum Sense goals must be incompatible with LookForWaste goal Our Clean goal will be created once our previously defined belief clean creates an event Our clean goal mus
30. setBatteryLevel batteryLevel out printlin Charging capa getBatteryLevel rplan waitFor 100 get capa agent setPriority KeepBatteryLoaded class 0 0 capa agent setPriority LookForWaste class 1 0 capa agent setPriority Clean class 0 0 return ret Figure 42 KeepBatteryLoadedPlan java Definition For goal deliberation process we must control our agent battery level continuously If battery level drops under a defined threshold we must change goals priorities to activate designed goals according to our created deliberative coherence model Open the SensePlan java file and update the body of the plan as in Fig 43 PlanBody public IFuture lt Void gt body final Future lt Void gt ret new Future lt Void gt Room room Room getInstance double dirtiness room getDirtiness capa getxPos capa getyPos double batteryLevel capa getBatteryLevel System out printin Dirtiness dirtiness BateryLevel batteryLevel capa setSensedDirt dirtiness if dirtiness gt 1 capa agent setPriority Clean class 1 0 capa agent setPriority LookForWaste class 0 0 if batteryLevel lt 0 4 capa agent setPriority Clean class 0 0 capa agent setPriority LookForWaste class 0 0 capa agent setPriority KeepBatteryLoaded class 1 0 return ret Figure 43 Updated SensePlan body Open the CleanerDeCo java file and add the following plan headings
31. t be unique and must inhibit the other Clean goals that are dispatched As can be done in BDIAgent we can use deliberation field of the Goal annotation to define uniqueness Add the following Clean class definition to CleanerDeCo java file deliberation Deliberation inhibits Clean class relations Relations facilitations Goal unique true Facilitate units Vacuum class degree 1 Facilitate units Sense class degree 1 incompatibles Incompatible unit LookForWaste class degree 1 priority 0 0 public class Clean GoalCreationCondition events clean public Clean public int hashCode final int prime 31 int result 1 result prime result getOuterType hashCode result prime result 0 return result public boolean equals Object obj boolean ret false if obj instanceof Clean Clean other Clean obj ret getOuterType equals other getOuterType return ret private CleanerDeCo getOuterType return CleanerDeCo this Figure 33 Clean Goal Definition Create a new java file and name it as CleanPlan java Our CleanPlan body will only dispatch Vacuum goal as a sub goal Add the following code to CleanPlan java file package deco jadex examples cleaner import jadex commons future ExceptionDelegationResultListener import jadex commons future Future imp
32. t d r dx d double ry r lt d r dy d capa setxPos posX rx capa setyPos posYtry rplan waitFor 100 addResultListener new DelegationResultListener lt Void gt ret J System out printin Cleaner moved to position X capa getxPos Y capa getyPos return ret Figure 14 MovePlan oneStepToTarget Function Definition Add the following necessary functions to the MovePlan public double getDistance double x double y double tarxX double tarY return Math sqrt y tarY y tarY x tarX x tarX public boolean isNear double x double y double tarx double tarY return getDistance x y tarX tarY lt 0 01 Figure 15 Other Necessary Functions Now we completed the MovePlan implementation Close the MovePlan java and open the CleanerDeCo Java file Our second goal is the MoveAround goal MoveAround goal must be facilitated by Move goal according to our deliberative coherence model As we mentioned earlier we define relations with ReLation annotation at relations field of the goal and we define priority of the goal at priority field We add facilitation relationship into the facilitations field of the ReLations Fig 16 shows FaciLlitate annotation usage This type of usage means that Move java goal facilitates MoveAround java goal and this is what we want to create with deliberative coherence model Add the following MoveAround goal definition to
33. t x and y coordinates from Move goal and tries to lead the agent there To satisfy this agent must change its current location in every step of execution through to the target position Add the following moveToDestinationFunction after the body of the MovePlan protected IFuture lt Void gt moveToDestination final Future lt Void gt ret new Future lt Void gt double destX goal getXTarPos double destY goal getYTarPos double posX capa getxPos double posY capa getyPos if isNear posX posY destX destyY oneStepToTarget addResultListener new DelegationResultListener lt Void gt ret public void customResultAvailable Void result moveToDestination addResultListener new DelegationResultListener lt Void gt ret ret setResult null return ret Figure 13 MovePlan moveToDestination Function Definition At every execution step agent will get closer to the target position until it reaches to the target Add the following oneStepToTarget function after the moveToDestination function protected IFuture lt Void gt oneStepToTarget final Future lt Void gt ret new Future lt Void gt double destX goal getXTarPos double destY goal getYTarPos double posX capa getxPos double posY capa getyPos double d getDistance posX posY destX destyY double r 0 00004 100 double dx destX posX double dy destY posyY double rx r l
34. tery level of the agent above a level Additionally we will add ReturnHome goal to return the cleaner to the battery station Our KeepBattery goal will be facilitated by Return Home goal and will be incompatible with LookForWaste and Clean goals Our ReturnHome goal will be facilitated by Move goal Desirability 05 Desirability 05 Desirability 05 Figure 37 DeCo Model Activations The appearance of the DeCo model activations LookForWaste priority is 1 Clean priority is 0 and KeepBatteryLoaded priority is 0 left LookForWaste priority is 0 Clean priority is 1 and KeepBatteryLoaded priority is 0 right LookForWaste priority 3 Implementation Open CleanerDeCo java file and add batteryLevel belief to store the battery level of the agent and recharge dynamic belief to define creation condition of the MaintainBatteryLoaded goal Belief private double batteryLevel 1 0 Belief updaterate 100 private boolean recharge batteryLevel lt 0 4 public double getBatteryLevel return batteryLevel public void setBatteryLevel double batteryLevel this batteryLevel batteryLevel Figure 38 batteryLevel and recharge Beliefs Add the ReturnHome goal to our agent definition file Since the ReturnHome goal is facilitated by Move goal define a facilitation relationship between them Goal relations Relations facilitations Facilitate units Move cl
35. ts priority will be assigned as internally JadexBDIv3 supports GoalCreationCondition GoalTargetCondition GoaLDropCondition GoalContextCondition method conditions for a goal DeCoAgent Library supports all of these abilities except G oalContextCondition Because DeCoAgent takes the context handling Add the Move class in to the agent definition as an inner class Goal public class Move private double xTargetPos private double yTargetPos public Move double x double y xTarget Pos x ylargetPos y GoalTargetCondition events xPos yPos public boolean checkTarget boolean ret false double distance Math sqrt xPos xTargetPos xPos xTargetPos yPos yTargetPos yPos yTargetPos if distance lt 0 01 ret true return ret public double getXTarPos return xTargetPos public double getYTarPos return yTargetPos Figure 11 Move Goal Definition Goal denotes the fact that an agent commits itself to a certain objective and maybe tries all the possibilities to achieve its goal Plans support goals by achieving agent to some state of affair A plan defines two aspects In the head of the plan i e in its PLan annotation meta information about the plan is defined This means that in the plan head several properties of the plan can be specified e g the circumstances under which it is activated and its importance in relation to other plans The bod
36. y is 0 Clean priority is 1 right is depicted in Fig 29 Desirability 05 Desirability 05 Figure 29 The Appearance of the DeCo Model Activations 3 Implementation Open CleanerDeCo java file and add a clean dynamic belief which is updated on every rate update This belief will be our creation condition of the clean goal Belief updaterate 200 private boolean clean sensedDirt gt 1 Figure 30 clean Dynamic Belief Definition Add vacuum goal as an empty goal which is shown in Fig 30 to the agent definition file Goal public class Vacuum public Vacuum Figure 31 Vacuum Goal Definition Create a new java document named as VacuumPlan java In VacuumPlan body we get the current sensation of the agent and check if it is above the dirtiness threshold If so agent will vacuum the room with its synchronized function Vacuum When the dirtiness of the room is removed cleaning goal will be accomplished After accomplishing the cleaning goal we must reconfigure our goals priorities We set MoveAround goal priority to 1 and Clean goal priority to 0 package deco jadex examples cleaner import jadex commons future Future import jadex commons future IFuture import deco jadex bdiv3 annotation Plan import deco jadex bdiv3 annotation PlanAPI import deco jadex bdiv3 annotation PlanBody import deco jadex bdiv3 annotation PlanCapability import deco jadex bdiv3 runtime IPlan impo
37. y of a plan contains the concrete instruction that should be carried out Plans can be defined as methods inner classes or classes Now create a new Java class and name it as MovePlan java A plan must start with P Lan annotation in Jadex PLanCapability defines the API which is used this plan PLanAPI defines the plan api of the library and the PLanReason defines the aim of the current plan Every plan must have a body Each plan body must be annotated with PLanBody Plan body defines the path of an agent that must follow for achieving its goal Add the following MovePlan definition to the MovePlan java file package deco jadex examples cleaner import deco jadex bdiv3 annotation Plan import deco jadex bdiv3 annotation PlanAPI import deco jadex bdiv3 annotation PlanBody import deco jadex bdiv3 annotation PlanCapability import deco jadex bdiv3 annotation PlanReason import deco jadex bdiv3 runtime IPlan import deco jadex examples cleaner CleanerDeCo Move import jadex commons future DelegationResultListener import jadex commons future Future import jadex commons future IFuture Plan public class MovePlan PlanCapability protected CleanerDeCo capa PlanAPI protected IPlan rplan PlanReason protected Move goal public MovePlan kl PlanBody public IFuture lt Void gt body return moveToDestination Figure 12 Move Plan Definition Our move plan takes the targe
Download Pdf Manuals
Related Search