burlap.behavior.singleagent.planning

Class QComputablePlanner.QComputablePlannerHelper

java.lang.Object

burlap.behavior.singleagent.planning.QComputablePlanner.QComputablePlannerHelper

Enclosing interface:

QComputablePlanner

public static class QComputablePlanner.QComputablePlannerHelper
extends java.lang.Object

A class of helper static methods that may be commonly used code that uses a QComputable planner. In particular, methods for computing the value function of a state, given the Q-values

Constructor Summary

Constructors

Constructor and Description
QComputablePlanner.QComputablePlannerHelper()

Method Summary

Methods

Modifier and Type	Method and Description
static double	getOptimalValue(QComputablePlanner planner, State s) Returns the optimal state value function for a state given a QComputablePlanner.
static double	getOptimalValue(QComputablePlanner planner, State s, TerminalFunction tf) Returns the optimal state value for a state given a QComputablePlanner.
static double	getPolicyValue(QComputablePlanner planner, State s, Policy p) Returns the state value under a given policy for a state and QComputablePlanner.
static double	getPolicyValue(QComputablePlanner planner, State s, Policy p, TerminalFunction tf) Returns the state value under a given policy for a state and QComputablePlanner.

Methods inherited from class java.lang.Object

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

Constructor Detail

QComputablePlanner.QComputablePlannerHelper

public QComputablePlanner.QComputablePlannerHelper()

Method Detail

getOptimalValue

public static double getOptimalValue(QComputablePlanner planner,
                     State s)

Returns the optimal state value function for a state given a QComputablePlanner. The optimal value is the max Q-value. If no actions are permissible in the input state, then zero is returned.

Parameters:

planner - the QComputablePlanner capable of producing Q-values.

s - the query State for which the value should be returned.

Returns:

the max Q-value for all possible Q-values in the state.

getOptimalValue

public static double getOptimalValue(QComputablePlanner planner,
                     State s,
                     TerminalFunction tf)

Returns the optimal state value for a state given a QComputablePlanner. The optimal value is the max Q-value. If no actions are permissible in the input state or the input state is a terminal state, then zero is returned.

Parameters:

planner - the QComputablePlanner capable of producing Q-values.

s - the query State for which the value should be returned.

tf - a terminal function.

Returns:

the max Q-value for all possible Q-values in the state or zero if there are not permissible actions or if the state is a terminal state.

getPolicyValue

public static double getPolicyValue(QComputablePlanner planner,
                    State s,
                    Policy p)

Returns the state value under a given policy for a state and QComputablePlanner. The value is the expected Q-value under the input policy action distribution. If no actions are permissible in the input state, then zero is returned.

Parameters:

planner - the QComputablePlanner capable of producing Q-values.

s - the query State for which the value should be returned.

p - the policy defining the action distribution.

Returns:

the expected Q-value under the input policy action distribution

getPolicyValue

public static double getPolicyValue(QComputablePlanner planner,
                    State s,
                    Policy p,
                    TerminalFunction tf)

Returns the state value under a given policy for a state and QComputablePlanner. The value is the expected Q-value under the input policy action distribution. If no actions are permissible in the input state, then zero is returned.

Parameters:

planner - the QComputablePlanner capable of producing Q-values.

s - the query State for which the value should be returned.

p - the policy defining the action distribution.

tf - a terminal function.

Returns:

the expected Q-value under the input policy action distribution or zero if there are not permissible actions or if the state is a terminal state.