aima.core.search.framework

Class ProblemSolvingAgent

java.lang.Object

aima.core.agent.impl.AbstractAgent

aima.core.search.framework.ProblemSolvingAgent

All Implemented Interfaces:

Agent, EnvironmentObject

Direct Known Subclasses:

MapAgent

public abstract class ProblemSolvingAgent
extends AbstractAgent

Modified copy of class aima.search.framework.SimpleProblemSolvingAgent which can be used for online search, too. Here, attribute plan (original: seq) is protected. Static pseudo code variable state is used in a more general sense including world state as well as agent state aspects. This allows the agent to change the plan, if unexpected percepts are observed. In the concrete java code, state corresponds with the agent instance itself (this).

 
 function PROBLEM-SOLVING-AGENT(percept) returns an action
   inputs: percept, a percept
   static: state, some description of current agent and world state
           
   state <- UPDATE-STATE(state, percept)
   while (state.plan is empty) do
     goal <- FORMULATE-GOAL(state)
     if (goal != null) then
       problem    <- FORMULATE-PROBLEM(state, goal)
       state.plan <- SEARCH(problem)
       if (state.plan is empty and !tryWithAnotherGoal()) then
         add NO_OP to plan         // failure
     else
       add NO_OP to plan           // success
   action <- FIRST(state.plan)
   plan <- REST(state.plan)
   return action
 
 

Author:

Ruediger Lunde

Field Summary

Fields

Modifier and Type	Field and Description
protected java.util.List<Action>	plan Plan, an action sequence, initially empty.

Fields inherited from class aima.core.agent.impl.AbstractAgent

program

Constructor Summary

Constructors

Constructor and Description
ProblemSolvingAgent()

Method Summary

Modifier and Type	Method and Description
Action	execute(Percept p) Template method, which corresponds to pseudo code function PROBLEM-SOLVING-AGENT(percept).
protected abstract java.lang.Object	formulateGoal() Primitive operation, responsible for goal generation.
protected abstract Problem	formulateProblem(java.lang.Object goal) Primitive operation, responsible for search problem generation.
protected abstract java.util.List<Action>	search(Problem problem) Primitive operation, responsible for the generation of an action list (plan) for the given search problem.
protected boolean	tryWithAnotherGoal() Primitive operation, which decides after a search for a plan failed, whether to stop the whole task with a failure, or to go on with formulating another goal.
protected abstract java.lang.Object	updateState(Percept p) Primitive operation, responsible for updating the state of the agent with respect to latest feedback from the world.

Methods inherited from class aima.core.agent.impl.AbstractAgent

isAlive, setAlive

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

plan

protected java.util.List<Action> plan

Plan, an action sequence, initially empty.

Constructor Detail

ProblemSolvingAgent

public ProblemSolvingAgent()

Method Detail

execute

public Action execute(Percept p)

Template method, which corresponds to pseudo code function PROBLEM-SOLVING-AGENT(percept).

Specified by:

execute in interface Agent

Overrides:

execute in class AbstractAgent

Parameters:

p - The current percept of a sequence perceived by the Agent.

Returns:

an action

tryWithAnotherGoal

protected boolean tryWithAnotherGoal()

Primitive operation, which decides after a search for a plan failed, whether to stop the whole task with a failure, or to go on with formulating another goal. This implementation always returns false. If the agent defines local goals to reach an externally specified global goal, it might be interesting, not to stop when the first local goal turns out to be unreachable.

updateState

protected abstract java.lang.Object updateState(Percept p)

Primitive operation, responsible for updating the state of the agent with respect to latest feedback from the world. In this version, implementations have access to the agent's current goal and plan, so they can modify them if needed. For example, if the plan didn't work because the model of the world proved to be wrong, implementations could update the model and also clear the plan.

formulateGoal

protected abstract java.lang.Object formulateGoal()

Primitive operation, responsible for goal generation. In this version, implementations are allowed to return null to indicate that the agent has finished the job an should die. Implementations can access the current goal (which is a possibly modified version of the last formulated goal). This might be useful in situations in which plan execution has failed.

formulateProblem

protected abstract Problem formulateProblem(java.lang.Object goal)

Primitive operation, responsible for search problem generation.

search

protected abstract java.util.List<Action> search(Problem problem)

Primitive operation, responsible for the generation of an action list (plan) for the given search problem.