burlap.behavior.singleagent.learning.actorcritic.critics

Class TimeIndexedTDLambda

java.lang.Object

burlap.behavior.singleagent.learning.actorcritic.critics.TDLambda

burlap.behavior.singleagent.learning.actorcritic.critics.TimeIndexedTDLambda

All Implemented Interfaces:

Critic, ValueFunction

public class TimeIndexedTDLambda
extends TDLambda

An implementation of TDLambda that can be used as a critic for ActorCritic algorithms [1], except that this class treats states at different depths as unique states. In general the typical TDLambda method is recommend unless a special Actor object that exploits the time information is to be used as well.

1. Barto, Andrew G., Steven J. Bradtke, and Satinder P. Singh. "Learning to act using real-time dynamic programming." Artificial Intelligence 72.1 (1995): 81-138.

Author:

James MacGlashan

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	TimeIndexedTDLambda.StateTimeElibilityTrace Extends the standard TDLambda.StateEligibilityTrace to include time/depth information.

Nested classes/interfaces inherited from class burlap.behavior.singleagent.learning.actorcritic.critics.TDLambda

TDLambda.StateEligibilityTrace

Field Summary

Fields

Modifier and Type	Field and Description
protected int	curTime The current time index / depth of the current episode
protected int	maxEpisodeSize The maximum number of steps possible in an episode.
protected java.util.List<java.util.Map<HashableState,burlap.behavior.singleagent.learning.actorcritic.critics.TDLambda.VValue>>	vTIndex The time/depth indexed value function

Fields inherited from class burlap.behavior.singleagent.learning.actorcritic.critics.TDLambda

gamma, hashingFactory, lambda, learningRate, rf, tf, totalNumberOfSteps, traces, vIndex, vInitFunction

Constructor Summary

Constructors

Constructor and Description
TimeIndexedTDLambda(RewardFunction rf, TerminalFunction tf, double gamma, HashableStateFactory hashingFactory, double learningRate, double vinit, double lambda) Initializes the algorithm.
TimeIndexedTDLambda(RewardFunction rf, TerminalFunction tf, double gamma, HashableStateFactory hashingFactory, double learningRate, double vinit, double lambda, int maxEpisodeSize) Initializes the algorithm.
TimeIndexedTDLambda(RewardFunction rf, TerminalFunction tf, double gamma, HashableStateFactory hashingFactory, double learningRate, ValueFunctionInitialization vinit, double lambda, int maxEpisodeSize) Initializes the algorithm.

Method Summary

Methods

Modifier and Type	Method and Description
CritiqueResult	critiqueAndUpdate(State s, GroundedAction ga, State sprime) This method's implementation provides the critique for some specific instance of the behavior.
void	endEpisode() This method is called whenever a learning episode terminates
int	getCurTime() Returns the current time/depth of the current episodes
protected burlap.behavior.singleagent.learning.actorcritic.critics.TDLambda.VValue	getV(HashableState sh, int t) Returns the TDLambda.VValue object (storing the value) for a given hashed stated at the specified time/depth.
void	initializeEpisode(State s) This method is called whenever a new learning episode begins
void	resetData() Used to reset any data that was created/modified during learning so that learning can be begin anew.
void	setCurTime(int t) Sets the time/depth of the current episode.

Methods inherited from class burlap.behavior.singleagent.learning.actorcritic.critics.TDLambda

addNonDomainReferencedAction, getV, setLearningRate, setRewardFunction, value

Methods inherited from class java.lang.Object

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

Field Detail

vTIndex

protected java.util.List<java.util.Map<HashableState,burlap.behavior.singleagent.learning.actorcritic.critics.TDLambda.VValue>> vTIndex

The time/depth indexed value function

curTime

protected int curTime

The current time index / depth of the current episode

maxEpisodeSize

protected int maxEpisodeSize

The maximum number of steps possible in an episode.

Constructor Detail

TimeIndexedTDLambda

public TimeIndexedTDLambda(RewardFunction rf,
                   TerminalFunction tf,
                   double gamma,
                   HashableStateFactory hashingFactory,
                   double learningRate,
                   double vinit,
                   double lambda)

Initializes the algorithm.

Parameters:

rf - the reward function

tf - the terminal state function

gamma - the discount factor

hashingFactory - the state hashing factory to use for hashing states and performing equality checks.

learningRate - the learning rate that affects how quickly the estimated value function is adjusted.

vinit - a constant value function initialization value to use.

lambda - indicates the strength of eligibility traces. Use 1 for Monte-carlo-like traces and 0 for single step backups

TimeIndexedTDLambda

public TimeIndexedTDLambda(RewardFunction rf,
                   TerminalFunction tf,
                   double gamma,
                   HashableStateFactory hashingFactory,
                   double learningRate,
                   double vinit,
                   double lambda,
                   int maxEpisodeSize)

Initializes the algorithm.

Parameters:

rf - the reward function

tf - the terminal state function

gamma - the discount factor

hashingFactory - the state hashing factory to use for hashing states and performing equality checks.

learningRate - the learning rate that affects how quickly the estimated value function is adjusted.

vinit - a constant value function initialization value to use.

lambda - indicates the strength of eligibility traces. Use 1 for Monte-carlo-like traces and 0 for single step backups

maxEpisodeSize - the maximum number of steps possible in an episode

TimeIndexedTDLambda

public TimeIndexedTDLambda(RewardFunction rf,
                   TerminalFunction tf,
                   double gamma,
                   HashableStateFactory hashingFactory,
                   double learningRate,
                   ValueFunctionInitialization vinit,
                   double lambda,
                   int maxEpisodeSize)

Initializes the algorithm.

Parameters:

rf - the reward function

tf - the terminal state function

gamma - the discount factor

hashingFactory - the state hashing factory to use for hashing states and performing equality checks.

learningRate - the learning rate that affects how quickly the estimated value function is adjusted.

vinit - a method of initializing the value function for previously unvisited states.

lambda - indicates the strength of eligibility traces. Use 1 for Monte-carlo-like traces and 0 for single step backups

maxEpisodeSize - the maximum number of steps possible in an episode

Method Detail

getCurTime

public int getCurTime()

Returns the current time/depth of the current episodes

Returns:

the current time/depth of the current episodes

setCurTime

public void setCurTime(int t)

Sets the time/depth of the current episode.

Parameters:

t - the time/depth of the current episode.

initializeEpisode

public void initializeEpisode(State s)

Description copied from interface: Critic

This method is called whenever a new learning episode begins

Specified by:

initializeEpisode in interface Critic

Overrides:

initializeEpisode in class TDLambda

Parameters:

s - the initial state of the new learning episode

endEpisode

public void endEpisode()

Description copied from interface: Critic

This method is called whenever a learning episode terminates

Specified by:

endEpisode in interface Critic

Overrides:

endEpisode in class TDLambda

critiqueAndUpdate

public CritiqueResult critiqueAndUpdate(State s,
                               GroundedAction ga,
                               State sprime)

Description copied from interface: Critic

This method's implementation provides the critique for some specific instance of the behavior.

Specified by:

critiqueAndUpdate in interface Critic

Overrides:

critiqueAndUpdate in class TDLambda

Parameters:

s - an input state

ga - an action taken in s

sprime - the state the agent transitioned to for taking action ga in state s

Returns:

the critique of this behavior.

getV

protected burlap.behavior.singleagent.learning.actorcritic.critics.TDLambda.VValue getV(HashableState sh,
                                                                            int t)

Returns the TDLambda.VValue object (storing the value) for a given hashed stated at the specified time/depth.

Parameters:

sh - the hashed state for which the value should be returned.

t - the time/depth at which the state is visited

Returns:

the TDLambda.VValue object (storing the value) for a given hashed stated at the specified time/depth

resetData

public void resetData()

Description copied from interface: Critic

Used to reset any data that was created/modified during learning so that learning can be begin anew.

Specified by:

resetData in interface Critic

Overrides:

resetData in class TDLambda