burlap.behavior.singleagent.learning.modellearning.rmax

Class PotentialShapedRMax.RMaxPotential

java.lang.Object

burlap.behavior.singleagent.learning.modellearning.rmax.PotentialShapedRMax.RMaxPotential

All Implemented Interfaces:

PotentialFunction

Enclosing class:

PotentialShapedRMax

public static class PotentialShapedRMax.RMaxPotential
extends java.lang.Object
implements PotentialFunction

A potential function for vanilla RMax; all states have a potential value of R_max/(1-gamma)

Author:

James MacGlashan

Constructor Summary

Constructors

Constructor and Description
RMaxPotential(double vMax) Initializes using the given maximum value function value
RMaxPotential(double rMax, double gamma) Initializes for a given maximum reward and discount factor.

Method Summary

Modifier and Type	Method and Description
double	potentialValue(State s) Returns the reward potential from the given state.

Methods inherited from class java.lang.Object

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

Constructor Detail

RMaxPotential

public RMaxPotential(double rMax,
                     double gamma)

Initializes for a given maximum reward and discount factor. Sets potential for all states to rMax/(1-gamma)

Parameters:

rMax - the maximum possible reward

gamma - the discount factor.

RMaxPotential

public RMaxPotential(double vMax)

Initializes using the given maximum value function value

Parameters:

vMax - the maximum value function value

Method Detail

potentialValue

public double potentialValue(State s)

Description copied from interface: PotentialFunction

Returns the reward potential from the given state. Note: the potential function should always return 0 for terminal states.

Specified by:

potentialValue in interface PotentialFunction

Parameters:

s - the input state for which to get the reward potential.

Returns:

the reward potential from the given state.