burlap.domain.stochasticgames.gridgame

Class GridGameStandardMechanics

java.lang.Object

burlap.oomdp.stochasticgames.JointActionModel

burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics

public class GridGameStandardMechanics
extends JointActionModel

This class defines the standard transition dynamics for a grid game. By default, the dynamics set a 0.5 probability of agents being able to move through semi-walls. The semi-wall pass through probability, however, may be changed by a parameter to a constructor. If two agents adjacent agents try to move into each others locations, no one will move. If an agent tries to move into a cell that the other agent is leaving, then they both will move. If two agents try to move into the same location, then only one of them will randomly make it, while the other will be unmoved in their position.

Author:

James MacGlashan

Constructor Summary

Constructors

Constructor and Description
GridGameStandardMechanics(Domain d) Initializes the mechanics for the given domain and sets the semi-wall pass through probability to 0.5;
GridGameStandardMechanics(Domain d, double semiWallPassThroughProb) Initializes the mechanics for the given domain and sets the semi-wall pass through probability to semiWallPassThroughProb.

Method Summary

Methods

Modifier and Type	Method and Description
protected State	actionHelper(State s, JointAction ja) This method is what determines the state when JointAction ja is executed in State s.
protected boolean	agentsEqual(State s1, State s2) Returns true if the agent objects between these two states are equal
protected burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2	attemptedDelta(java.lang.String actionName) Returns the attempted change in position by the agent for the given action.
protected java.util.List<TransitionProbability>	combineDuplicateTransitionProbabilities(java.util.List<TransitionProbability> srcTPs) Iterates through a list of transition probability objects and combines any that refer to the same state
protected boolean	crossesWall(burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 p0, burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 delta, ObjectInstance w, boolean vertical) Returns true if the agent would cross a given wall instance given a movement attempt.
protected void	generateAllLocationSetsHelper(java.util.List<java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2Prob>> locOutcomes, int i, burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2[] locArray, double stackedProb, java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.LocationSetProb> discovered) This method will recursively generate all possible joint location outcomes for a list of possible outcomes for each agent
protected void	generateAllPossibleCollisionWinnerAssignments(java.util.List<java.util.List<java.lang.Integer>> collisionLists, int i, int[] assignment, java.util.List<java.util.List<java.lang.Integer>> discovered)
protected java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.LocationSetProb>	getAllLocationSets(java.util.List<java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2Prob>> locOutcomes) Takes a list of possible location outcomes for each agent and generates all joint location outcomes
protected java.util.List<java.util.Map<java.lang.Integer,java.lang.Integer>>	getAllPossibleCollisionWinnerAssignment(java.util.Map<java.lang.Integer,java.util.List<java.lang.Integer>> collissionSets)
protected java.util.Map<java.lang.Integer,java.util.List<java.lang.Integer>>	getColissionSets(java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> candidatePositions) Returns with whom each agent is in a collision competition for a cell.
protected burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2	getLocation(State s, java.lang.String agentName) Returns the x-y position of an agent stored in a Location2 object.
protected java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.LocationSetProb>	getPossibleCollisionOutcomes(java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> originalPositions, java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> desiredPositions)
protected java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2Prob>	getPossibleLocationsFromWallCollisions(State s, burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 p0, burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 delta, java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> agentNoOpLocs) Returns the list of possible outcome locations for a given start point and desired position change.
protected java.util.Map<java.lang.Integer,java.lang.Integer>	getWinningAgentMovements(java.util.Map<java.lang.Integer,java.util.List<java.lang.Integer>> collissionSets) Takes as input the set of collisions and randomly selects a winner
protected java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2>	resolveCollisions(java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> originalPositions, java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> desiredPositions) Resolves collisions that occur when two or more agents try to enter the same cell, in which case only one agent will make it into the position and the rest will stay in place
protected java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2>	resolvePositionSwaps(java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> originalPositions, java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> desiredPositions) Returns the position of each agent after accounting for collisions that are a result of agents trying to move into each others previous locations.
protected burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2	sampleBasicMovement(State s, burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 p0, burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 delta, java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> agentNoOpLocs) Returns a movement result of the agent.
protected boolean	sampleWallCollision(burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 p0, burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 delta, java.util.List<ObjectInstance> walls, boolean vertical) Return true if the agent is able to move in the desired location; false if the agent moves into a solid wall or if the agent randomly fails to move through a semi-wall that is in the way.
java.util.List<TransitionProbability>	transitionProbsFor(State s, JointAction ja) Returns the transition probabilities for applying the provided JointAction action in the given state.
protected int	wallCollision(burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 p0, burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 delta, java.util.List<ObjectInstance> walls, boolean vertical) Indicates whether there are any wall collisions.

Methods inherited from class burlap.oomdp.stochasticgames.JointActionModel

deterministicTransitionProbsFor, performJointAction

Methods inherited from class java.lang.Object

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

Constructor Detail

GridGameStandardMechanics

public GridGameStandardMechanics(Domain d)

Initializes the mechanics for the given domain and sets the semi-wall pass through probability to 0.5;

Parameters:

d - the domain object

GridGameStandardMechanics

public GridGameStandardMechanics(Domain d,
                         double semiWallPassThroughProb)

Initializes the mechanics for the given domain and sets the semi-wall pass through probability to semiWallPassThroughProb.

Parameters:

d - d the domain object

semiWallPassThroughProb - the probability that an agent will pass through a semi-wall.

Method Detail

transitionProbsFor

public java.util.List<TransitionProbability> transitionProbsFor(State s,
                                                       JointAction ja)

Description copied from class: JointActionModel

Returns the transition probabilities for applying the provided JointAction action in the given state. Transition probabilities are specified as list of TransitionProbability objects. The list is only required to contain transitions with non-zero probability.

Specified by:

transitionProbsFor in class JointActionModel

Parameters:

s - the state in which the joint action is performed

ja - the joint action performed

Returns:

a list of state TransitionProbability objects.

actionHelper

protected State actionHelper(State s,
                 JointAction ja)

Description copied from class: JointActionModel

This method is what determines the state when JointAction ja is executed in State s. The input state should be directly modified.

Specified by:

actionHelper in class JointActionModel

Parameters:

s - the state in which the joint action is performed.

ja - the joint action to be performed.

Returns:

the resulting state of applying this action.

getLocation

protected burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 getLocation(State s,
                                                                                     java.lang.String agentName)

Returns the x-y position of an agent stored in a Location2 object.

Parameters:

s - the state in which the agent exists

agentName - the name of the agent.

Returns:

a GridGameStandardMechanics.Location2 object containing the agents position in the world.

attemptedDelta

protected burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 attemptedDelta(java.lang.String actionName)

Returns the attempted change in position by the agent for the given action. For instance, if the action is north, it would result in an attempted position change of (0, +1).

Parameters:

actionName - the action taken.

Returns:

the attempted change in position for the given action.

resolvePositionSwaps

protected java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> resolvePositionSwaps(java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> originalPositions,
                                                                                                              java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> desiredPositions)

Returns the position of each agent after accounting for collisions that are a result of agents trying to move into each others previous locations.

Parameters:

originalPositions - the original position of the agents before their actions were taken.

desiredPositions - the new position the agents are trying to go into

Returns:

the positions of the agents accounting for collisions.

resolveCollisions

protected java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> resolveCollisions(java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> originalPositions,
                                                                                                           java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> desiredPositions)

Resolves collisions that occur when two or more agents try to enter the same cell, in which case only one agent will make it into the position and the rest will stay in place

Parameters:

originalPositions - the positions of the agents in the original state before their actions were taken

desiredPositions - the desired locations of the agents

Returns:

a list of the resulting positions having accounted for collisions.

getPossibleCollisionOutcomes

protected java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.LocationSetProb> getPossibleCollisionOutcomes(java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> originalPositions,
                                                                                                                            java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> desiredPositions)

getWinningAgentMovements

protected java.util.Map<java.lang.Integer,java.lang.Integer> getWinningAgentMovements(java.util.Map<java.lang.Integer,java.util.List<java.lang.Integer>> collissionSets)

Takes as input the set of collisions and randomly selects a winner

Parameters:

collissionSets - the set of collisions; maps from one agent index to a list of the agents with whom he is competing for a cell

Returns:

A map from each agent involved in a collision to the winning agent of that collision.

getColissionSets

protected java.util.Map<java.lang.Integer,java.util.List<java.lang.Integer>> getColissionSets(java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> candidatePositions)

Returns with whom each agent is in a collision competition for a cell.

Parameters:

candidatePositions - a list of the positions to which each agent would like to go; the Location2 in index 1 indicates the location the 1th index agent wants to go to.

Returns:

A map from the index of each agent to a list of the index of agents with whom there is competition for a cell

sampleBasicMovement

protected burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 sampleBasicMovement(State s,
                                                                                             burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 p0,
                                                                                             burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 delta,
                                                                                             java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> agentNoOpLocs)

Returns a movement result of the agent. If the agent tries to pass through a semi-wall, then it is randomly selected whether the agent succeeds or not. If the agent tries to move through a solid wall or to a location where there is another agent who is not moving, then no change occurs.

Parameters:

s - the state containing the agent

p0 - the initial position of the agent

delta - the desired change of position.

agentNoOpLocs - the locations occupied by agents who are not moving.

Returns:

The resulting location of this agents movement.

getPossibleLocationsFromWallCollisions

protected java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2Prob> getPossibleLocationsFromWallCollisions(State s,
                                                                                                                                    burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 p0,
                                                                                                                                    burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 delta,
                                                                                                                                    java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2> agentNoOpLocs)

Returns the list of possible outcome locations for a given start point and desired position change.

Parameters:

s - the state in which the changes would occur

p0 - the initial location

delta - the desired change in position

agentNoOpLocs - the locations occupied by agents who are not moving

Returns:

the list of possible outcome locations in which the agent could wind up

sampleWallCollision

protected boolean sampleWallCollision(burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 p0,
                          burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 delta,
                          java.util.List<ObjectInstance> walls,
                          boolean vertical)

Return true if the agent is able to move in the desired location; false if the agent moves into a solid wall or if the agent randomly fails to move through a semi-wall that is in the way.

Parameters:

p0 - the initial location of the agent

delta - the desired change in direction of the agent

walls - the list of walls in the world

vertical - whether the list of provided walls are vertical or horizontal walls

Returns:

true if the agent is able to move in the desired location; false otherwise

wallCollision

protected int wallCollision(burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 p0,
                burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 delta,
                java.util.List<ObjectInstance> walls,
                boolean vertical)

Indicates whether there are any wall collisions.

Parameters:

p0 - the original location

delta - the desired change in direction

walls - possible wall objects

vertical - true if the wall objects are vertical; false if they are horizontal

Returns:

0 if there is no collision with a wall, 1 if there is a collision with a solid wall, 2 if there is a potential collision with a semi-wall

crossesWall

protected boolean crossesWall(burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 p0,
                  burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2 delta,
                  ObjectInstance w,
                  boolean vertical)

Returns true if the agent would cross a given wall instance given a movement attempt.

Parameters:

p0 - the initial state of the agent

delta - the desired change in direction

w - wall instance to check

vertical - true if the wall is a vertical wall; false if the wall is a horizontal wall

Returns:

whether the agent's movement would cross a wall

getAllLocationSets

protected java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.LocationSetProb> getAllLocationSets(java.util.List<java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2Prob>> locOutcomes)

Takes a list of possible location outcomes for each agent and generates all joint location outcomes

Parameters:

locOutcomes - a list of possible location outcomes

Returns:

all joint location outcomes

generateAllLocationSetsHelper

protected void generateAllLocationSetsHelper(java.util.List<java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2Prob>> locOutcomes,
                                 int i,
                                 burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.Location2[] locArray,
                                 double stackedProb,
                                 java.util.List<burlap.domain.stochasticgames.gridgame.GridGameStandardMechanics.LocationSetProb> discovered)

This method will recursively generate all possible joint location outcomes for a list of possible outcomes for each agent

Parameters:

locOutcomes - the list of possible location outcomes for each agent

i - the index from which to generate possible outcomes in a depth-first manner

locArray - an array holding the current assignments for a given depth

stackedProb - the probability of this location set occurring

discovered - the discovered combinations thus far

getAllPossibleCollisionWinnerAssignment

protected java.util.List<java.util.Map<java.lang.Integer,java.lang.Integer>> getAllPossibleCollisionWinnerAssignment(java.util.Map<java.lang.Integer,java.util.List<java.lang.Integer>> collissionSets)

generateAllPossibleCollisionWinnerAssignments

protected void generateAllPossibleCollisionWinnerAssignments(java.util.List<java.util.List<java.lang.Integer>> collisionLists,
                                                 int i,
                                                 int[] assignment,
                                                 java.util.List<java.util.List<java.lang.Integer>> discovered)

combineDuplicateTransitionProbabilities

protected java.util.List<TransitionProbability> combineDuplicateTransitionProbabilities(java.util.List<TransitionProbability> srcTPs)

Iterates through a list of transition probability objects and combines any that refer to the same state

Parameters:

srcTPs - and initial list of transition probability objects

Returns:

an output list of transition probability objects in which any duplicate states have been mereged

agentsEqual

protected boolean agentsEqual(State s1,
                  State s2)

Returns true if the agent objects between these two states are equal

Parameters:

s1 - the first state

s2 - the second state

Returns:

true if the agent objects between these two states are equal