burlap.behavior.singleagent.planning.stochastic.montecarlo.uct

Class UCTActionNode

java.lang.Object

burlap.behavior.singleagent.planning.stochastic.montecarlo.uct.UCTActionNode

public class UCTActionNode
extends java.lang.Object

UCT Action node that stores relevant action statics necessary for UCT.

Author:

James MacGlashan

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	UCTActionNode.UCTActionConstructor A factory for generating UCTActionNode objects.

Field Summary

Fields

Modifier and Type	Field and Description
Action	action The action this action node wraps
int	n The number of of times this action node has been taken
java.util.Map<HashableState,java.util.List<UCTStateNode>>	successorStates The possible successor states.
double	sumReturn The sum return observed for this action node

Constructor Summary

Constructors

Constructor and Description
UCTActionNode(Action a) Generates a new action node for a given action.

Method Summary

Modifier and Type	Method and Description
void	addSuccessor(UCTStateNode node) Adds a successor node to the list of possible successors
double	averageReturn() Returns the average return
java.util.List<UCTStateNode>	getAllSuccessors() Returns a list of all successor nodes observed
boolean	referencesSuccessor(UCTStateNode node) Returns whether this action node has a observed a given successor state node in the past
void	update(double sampledReturn) Updates the node statistics with a sample return

Methods inherited from class java.lang.Object

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

Field Detail

action

public Action action

The action this action node wraps

sumReturn

public double sumReturn

The sum return observed for this action node

n

public int n

The number of of times this action node has been taken

successorStates

public java.util.Map<HashableState,java.util.List<UCTStateNode>> successorStates

The possible successor states. Stores a list of nodes for the same outcome state since options may reach the same outcome state after a different number steps causing a further depth in the tree.

Constructor Detail

UCTActionNode

public UCTActionNode(Action a)

Generates a new action node for a given action. All statistics are initialized to 0.

Parameters:

a - the action this node wraps.

Method Detail

averageReturn

public double averageReturn()

Returns the average return

Returns:

the average return

update

public void update(double sampledReturn)

Updates the node statistics with a sample return

Parameters:

sampledReturn - the sample return observed

addSuccessor

public void addSuccessor(UCTStateNode node)

Adds a successor node to the list of possible successors

Parameters:

node - the uct successor node

referencesSuccessor

public boolean referencesSuccessor(UCTStateNode node)

Returns whether this action node has a observed a given successor state node in the past

Parameters:

node - the node which is checked to be in the current successor states

Returns:

true if this node contains in its observed successors the input state node

getAllSuccessors

public java.util.List<UCTStateNode> getAllSuccessors()

Returns a list of all successor nodes observed

Returns:

a list of all successor nodes observed