Achieving Agent Adaptability through Componentbased Mechanisms

1
Achieving Agent Adaptability through
Component-based Mechanisms

• Yizhi Zhao
• 2007-01

2
Contents

• What is an agent?
• Where are agents used?
• How are agent-based system built?
• How the adaptation taken into account?
• What are the existing techniques to implement
  adaptation?
• What is the Agent adaptation experiment?
• Conclusion and Future work

3
Contents

• What is an agent?
• Where are agents used?
• How are agent-based system built?
• How the adaptation taken into account?
• What are the existing techniques to implement
  adaptation?
• What is the Agent adaptation experiment?
• Conclusion and Future work

4
An example of agent

• Considerable work of a travel agent
• Understand what the user wants
• Actually act
• Behave without supervision
• Use experience

Communicative
Capable
Autonomous
Adaptive
5
Agent Mindset

• Agent is
• An encapsulated computer system
• Capable of flexible and autonomous action
• In dynamic and unpredictable environment
• Nicholas R. Jennings, 2000

6
Contents

• What is an agent?
• Where are agents used?
• How are agent-based system built?
• How the adaptation taken into account?
• What are the existing techniques to implement
  adaptation?
• What is the Agent adaptation experiment?
• Conclusion and Future work

7
Agent-based applications

• Application domains
• Systems and network management
• Mobile access and management
• Mail and messaging
• Information access and management
• Electronic Commerce
• Adaptive User Interfaces

8
Contents

• What is an agent?
• Where are agents used?
• How are agent-based system built?
• How the adaptation taken into account?
• What are the existing techniques to implement
  adaptation?
• What is the Agent adaptation experiment?
• Conclusion and Future work

9
Agent Architecture

• BDI model

10
Contents

• What is an agent?
• Where are agents used?
• How are agent-based system built?
• How the adaptation taken into account?
• What are the existing techniques to implement
  adaptation?
• What is the Agent adaptation experiment?
• Conclusion and Future work

11
Look at adaptation

• The scenario is
• Agent move around different contexts
• Notify, identify and reflect to the changes
• Need new code in the new context
• Adapt itself to the new environment

12
Adaptation Mechanism

• Self-adaptation
• Core agent

13
Contents

• What is an agent?
• Where are agents used?
• How are agent-based system built?
• How the adaptation taken into account?
• What are the existing techniques to implement
  adaptation?
• What is the Agent adaptation experiment?
• Conclusion and Future work

14
Look at component-based techniques

• Common object request broker architecture

• DCOM

4
3
2
1
15
Look at communication techniques

• Simple object access protocol
• Platform-independent manner
• Use XML and HTTP to access services, objects and
  servers

16
Contents

• What is an agent?
• Where are agents used?
• How are agent-based system built?
• How the adaptation taken into account?
• What are the existing techniques to implement
  adaptation?
• What is the Agent adaptation experiment?
• Conclusion and Future work

17
Agent adaptation experiment

• Elements and process
• Different sub-domains
• A client application to generate agents
• Pronouncers are responsible for managing and
  communicating

agent
18
Main challenges

• How does pronouncer find the difference of coming
  agents
• Avoid duplication of agent delivery
• Combine binary components without pre-defined
  references
• Exchange messages in distributed environment

19
Possible techniques to solve problems

• A logic base - The agents need learn
• What they should look for
• Where they should proceed towards

20
Agent migration

• Not stay on separate machines and communicate
  over a common network

• Traditional network architecture

21
Starting migration process

• Migrate both code and data from the sender host
  and resume the execution at the receiver host

go(tcp//ce.et. tudelft - yizhi 4040,
runAtRemote)
22
And resuming execution

• Agents code must be linked to the code of the
  already running agent host
• Dynamic class loading mechanism (at runtime)
• The relative rules
• Check if loaded before
• System class loader search from CLASSPATH
• Loading from system JAR file
• Load remote class file over the network by
  user-defined class loader

23
Class loader look ups classes
private directory public MyClassLoader(String
dir) directory dir public synchronized
Class loadClass(String name) Class c
findLoadedClass(name) if (c ! null) return
c try c findSystemClass(name) return
c catch (ClassNotFoundException e)
try byte data getClassData(directory,
name) return defineClass(name, data, 0,
data.length()) catch (IOException e) throw
new ClassNotFoundException() byte0
getClassData( . ..) . . // omitted for
brevity
Where is the class file
Whether the class has been loaded
The run-time class object from the class file
24
Binary perspective by reflection

• Reflection allows us to look inside a java object
  at runtime

Protected void startAgent (Agent theAgent,
String methodName) Class agentClass
theAgent.getClass() Method method
agentClass.getMethod(methodName, new
Class0) method.invoke(methodName, new
Object0)
The class name of the given agent object
Ask this class for a name methodName
Invoke the method
25
Another adaptive strategy

• Stated before, convert binary into object and
  examine the object through its name
• Just-In-Time customization

26
Contents

• What is an agent?
• Where are agents used?
• How are agent-based system built?
• How the adaptation taken into account?
• What are the existing techniques to implement
  adaptation?
• What is the Agent adaptation experiment?
• Conclusion and Future work

27
Conclusion

• Agent is a different way to solve complexity
• Agent to achieve it tasks through migration
• We look at the view of agent coming into the new
  environment
• Achieve adaptability using self-adaptation
  mechanism
• Respective techniques (CBD, Compiler, SOAP) to do
  an experiment

28
Future work

• We intend agents and remote host (pronouncers) to
  communicate with each other
• A user-defined class loader working with
  reflection mechanism
• Construct the experiment to see if
  self-adaptation could achieve agents tasks

29
Thank You and Questions !