dlife.rl
Class ActionSelectionPolicy

java.lang.Object
  dlife.rl.ActionSelectionPolicy

All Implemented Interfaces:

Serializable

Direct Known Subclasses:

EpsilonGreedyActionSelectionPolicy, ThresholdActionSelectionPolicy

public abstract class ActionSelectionPolicy
extends Object
implements Serializable

Abstract base class for objects that set the policy for action selection in a TD-based reinforcement learning agent. Objects implementing this interface are able to use a QTable and the current State to select an action to be performed.

Version:

Apr 26, 2011

Author:

Grant Braught, Dickinson College

See Also:

Serialized Form

Field Summary

protected QTable

qTable

Constructor Summary

ActionSelectionPolicy(QTable q)
Construct a new ActionSelectionPolicy that will be used to select actions based on the provided Q-Table.

Method Summary

abstract Action

getAction(ArrayList<Action> actions, State s) Choose one of the actions to take from the given State using the QTable provided to the constructor.

Methods inherited from class java.lang.Object

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

Field Detail

qTable

protected QTable qTable

Constructor Detail

ActionSelectionPolicy

public ActionSelectionPolicy(QTable q)

Construct a new ActionSelectionPolicy that will be used to select actions based on the provided Q-Table. The table itself will be available to sub-classes through the protected field.

Parameters:

q - the Q-Table.

Method Detail

getAction

public abstract Action getAction(ArrayList<Action> actions,
                                 State s)

Choose one of the actions to take from the given State using the QTable provided to the constructor. Implementations of this interface can choose different selection algorithms for different balances between exploitation and exploration (e.g. greedy, epsilon-greedy, soft-max, etc...).

Parameters:

actions - the list of Actions to choose from.

s - the current State.

Returns:

the Action to be taken.