burlap.behavior.singleagent.learnfromdemo.mlirl.support

Class QGradientPlannerFactory.DifferentiableVIFactory

java.lang.Object

burlap.behavior.singleagent.learnfromdemo.mlirl.support.QGradientPlannerFactory.DifferentiableVIFactory

All Implemented Interfaces:

QGradientPlannerFactory

Enclosing interface:

QGradientPlannerFactory

public static class QGradientPlannerFactory.DifferentiableVIFactory
extends java.lang.Object
implements QGradientPlannerFactory

A DifferentiableVI factory.

Nested Class Summary

Nested classes/interfaces inherited from interface burlap.behavior.singleagent.learnfromdemo.mlirl.support.QGradientPlannerFactory

QGradientPlannerFactory.DifferentiableVIFactory

Field Summary

Fields

Modifier and Type	Field and Description
protected HashableStateFactory	hashingFactory The HashableStateFactory used by the valueFunction.
protected double	maxDelta The value function change threshold to stop VI.
protected int	maxIterations The maximum allowed number of VI iterations.
protected TerminalFunction	tf The terminal function that the valueFunction uses.

Constructor Summary

Constructors

Constructor and Description
QGradientPlannerFactory.DifferentiableVIFactory(HashableStateFactory hashingFactory) Initializes the factory with the given HashableStateFactory.
QGradientPlannerFactory.DifferentiableVIFactory(HashableStateFactory hashingFactory, TerminalFunction tf, double maxDelta, int maxIterations) Initializes.

Method Summary

Methods

Modifier and Type	Method and Description
QGradientPlanner	generateDifferentiablePlannerForRequest(MLIRLRequest request) Returns a QGradientPlanner for an MLIRLRequest object's domain, reward function, discount factor, and Boltzmann beta parameter.

Methods inherited from class java.lang.Object

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

Field Detail

hashingFactory

protected HashableStateFactory hashingFactory

The HashableStateFactory used by the valueFunction.

maxDelta

protected double maxDelta

The value function change threshold to stop VI. Default is 0.01.

maxIterations

protected int maxIterations

The maximum allowed number of VI iterations. Default is 500.

tf

protected TerminalFunction tf

The terminal function that the valueFunction uses. Default is a a NullTermination.

Constructor Detail

QGradientPlannerFactory.DifferentiableVIFactory

public QGradientPlannerFactory.DifferentiableVIFactory(HashableStateFactory hashingFactory)

Initializes the factory with the given HashableStateFactory. The terminal function will be defaulted to a NullTermination; value function change threshold to 0.01; and the max VI iterations to 500.

Parameters:

hashingFactory - the HashableStateFactory to use for planning.

QGradientPlannerFactory.DifferentiableVIFactory

public QGradientPlannerFactory.DifferentiableVIFactory(HashableStateFactory hashingFactory,
                                               TerminalFunction tf,
                                               double maxDelta,
                                               int maxIterations)

Initializes.

Parameters:

hashingFactory - the HashableStateFactory to use for planning.

tf - The terminal function that the generated planners use.

maxDelta - The value function change threshold to stop VI.

maxIterations - The maximum allowed number of VI iterations

Method Detail

generateDifferentiablePlannerForRequest

public QGradientPlanner generateDifferentiablePlannerForRequest(MLIRLRequest request)

Description copied from interface: QGradientPlannerFactory

Returns a QGradientPlanner for an MLIRLRequest object's domain, reward function, discount factor, and Boltzmann beta parameter.

Specified by:

generateDifferentiablePlannerForRequest in interface QGradientPlannerFactory

Parameters:

request - the request defining the problem the valueFunction should solve.

Returns:

a QGradientPlanner instance.