burlap.behavior.singleagent.learning

Class GoalBasedRF

java.lang.Object

burlap.behavior.singleagent.learning.GoalBasedRF

All Implemented Interfaces:

RewardFunction

public class GoalBasedRF
extends java.lang.Object
implements RewardFunction

A reward function implementation designed around goal conditions that are specified by a StateConditionTest object. When the agent transition to a state marked as a goal state, it returns a goal reward. Otherwise a default reward is returned.

Author:

James MacGlashan

Field Summary

Fields

Modifier and Type	Field and Description
protected double	defaultReward
protected StateConditionTest	gc
protected double	goalReward

Constructor Summary

Constructors

Constructor and Description
GoalBasedRF(StateConditionTest gc) Initializes with transitions to goal states returning a reward of 1 and all others returning 0
GoalBasedRF(StateConditionTest gc, double goalReward) Initializes with transitions to goal states returning the given reward and all others returning 0.
GoalBasedRF(StateConditionTest gc, double goalReward, double defaultReward) Initializes with transitions to goal states returning the given reward and all others returning 0.
GoalBasedRF(TerminalFunction tf) Initializes with transitions to goal states, indicated by the terminal function, returning a reward of 1 and all others returning 0
GoalBasedRF(TerminalFunction tf, double goalReward) Initializes with transitions to goal states, indicated by the terminal function, returning the given reward and all others returning 0.
GoalBasedRF(TerminalFunction tf, double goalReward, double defaultReward) Initializes with transitions to goal states, indicated by the terminal function, returning the given reward and all others returning 0.

Method Summary

Methods

Modifier and Type	Method and Description
StateConditionTest	getGoalCondition() Returns the goal condition for this reward function.
double	reward(State s, GroundedAction a, State sprime) Returns the reward received when action a is executed in state s and the agent transitions to state sprime.

Methods inherited from class java.lang.Object

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

Field Detail

gc

protected StateConditionTest gc

goalReward

protected double goalReward

defaultReward

protected double defaultReward

Constructor Detail

GoalBasedRF

public GoalBasedRF(StateConditionTest gc)

Initializes with transitions to goal states returning a reward of 1 and all others returning 0

Parameters:

gc - StateConditionTest object that specifies goal states.

GoalBasedRF

public GoalBasedRF(StateConditionTest gc,
           double goalReward)

Initializes with transitions to goal states returning the given reward and all others returning 0.

Parameters:

gc - StateConditionTest object that specifies goal states.

goalReward - the reward returned for transitions to goal states.

GoalBasedRF

public GoalBasedRF(StateConditionTest gc,
           double goalReward,
           double defaultReward)

Initializes with transitions to goal states returning the given reward and all others returning 0.

Parameters:

gc - StateConditionTest object that specifies goal states.

goalReward - the reward returned for transitions to goal states.

defaultReward - the default reward returned for all non-goal state transitions.

GoalBasedRF

public GoalBasedRF(TerminalFunction tf)

Initializes with transitions to goal states, indicated by the terminal function, returning a reward of 1 and all others returning 0

Parameters:

tf - TerminalFunction object that specifies goal states.

GoalBasedRF

public GoalBasedRF(TerminalFunction tf,
           double goalReward)

Initializes with transitions to goal states, indicated by the terminal function, returning the given reward and all others returning 0.

Parameters:

tf - TerminalFunction object that specifies goal states.

goalReward - the reward returned for transitions to goal states.

GoalBasedRF

public GoalBasedRF(TerminalFunction tf,
           double goalReward,
           double defaultReward)

Initializes with transitions to goal states, indicated by the terminal function, returning the given reward and all others returning 0.

Parameters:

tf - TerminalFunction object that specifies goal states.

goalReward - the reward returned for transitions to goal states.

defaultReward - the default reward returned for all non-goal state transitions.

Method Detail

getGoalCondition

public StateConditionTest getGoalCondition()

Returns the goal condition for this reward function.

Returns:

the goal condition for this reward function.

reward

public double reward(State s,
            GroundedAction a,
            State sprime)

Description copied from interface: RewardFunction

Returns the reward received when action a is executed in state s and the agent transitions to state sprime.

Specified by:

reward in interface RewardFunction

Parameters:

s - the state in which the action was executed

a - the action executed

sprime - the state to which the agent transitioned

Returns:

the reward received when action a is executed in state s and the agent transitions to state sprime.