burlap.behavior.singleagent.options.support

Interface DirectOptionTerminateMapper

public interface DirectOptionTerminateMapper

If an option deterministically terminates with a fixed number of steps, then it may be useful for an option to immediately transition from the state in which the option was initiated to the end terminal state, rather than having to simulate each step of execution. This interface provides a means for defining the transition dynamics, the number of steps and cumulative reward, so that an option can immediately take the agent to the next state without having to perform simulation. After implementing this interface, it can then be provided to any Option object to prevent the option from performing the simulation. If the options path and number of steps cannot be guaranteed in advance, then this interface should not be implemented.

Author:

James MacGlashan

Method Summary

Methods

Modifier and Type	Method and Description
State	generateOptionTerminalState(State s) Returns the termination state that will result from applying the option associated with this object in the given initiation state.
double	getCumulativeReward(State s, State sp, RewardFunction rf, double discount) Returns the cumulative discounted reward that would be received from applying the option associated with this object in the given initiation state.
int	getNumSteps(State s, State sp) Returns the number of steps that would have occurred for taking the option associated with this object from state s and terminating in state sp.

Method Detail

generateOptionTerminalState

State generateOptionTerminalState(State s)

Returns the termination state that will result from applying the option associated with this object in the given initiation state.

Parameters:

s - the initiation state in which the option associated with this object will be applied.

Returns:

the resulting terminal state of the option.

getNumSteps

int getNumSteps(State s,
              State sp)

Returns the number of steps that would have occurred for taking the option associated with this object from state s and terminating in state sp.

Parameters:

s - the initiation state in which the option associated with this object will be applied.

sp - the terminal state the option will reach

Returns:

the number of steps that would have occurred

getCumulativeReward

double getCumulativeReward(State s,
                         State sp,
                         RewardFunction rf,
                         double discount)

Returns the cumulative discounted reward that would be received from applying the option associated with this object in the given initiation state.

Parameters:

s - the initiation state in which the option associated with this object will be applied.

sp - the terminal state the option will reach

rf - the reward function being used

discount - the discount factor

Returns:

the cumulative discounted reward