burlap.behavior.singleagent.planning.deterministic.informed.Heuristic Java Examples

@Test
public void testAStar() {
	GridWorldState initialState = new GridWorldState(new GridAgent(0, 0), new GridLocation(10, 10, 0, "loc0"));
	
	Heuristic mdistHeuristic = new Heuristic() {
		
		@Override
		public double h(State s) {

			GridAgent agent = ((GridWorldState)s).agent;
			GridLocation location = ((GridWorldState)s).locations.get(0);

			//get agent position
			int ax = agent.x;
			int ay = agent.y;
			
			//get location position
			int lx = location.x;
			int ly = location.y;
			
			//compute Manhattan distance
			double mdist = Math.abs(ax-lx) + Math.abs(ay-ly);
			
			return -mdist;
		}
	};
	
	//provide A* the heuristic as well as the reward function so that it can keep
	//track of the actual cost
	DeterministicPlanner planner = new AStar(domain, goalCondition,
		hashingFactory, mdistHeuristic);
	planner.planFromState(initialState);
	Policy p = new SDPlannerPolicy(planner);
	
	Episode analysis = PolicyUtils.rollout(p, initialState, domain.getModel());
	this.evaluateEpisode(analysis, true);
}
 

public void AStarExample(String outputPath){

		Heuristic mdistHeuristic = new Heuristic() {

			public double h(State s) {
				GridAgent a = ((GridWorldState)s).agent;
				double mdist = Math.abs(a.x-10) + Math.abs(a.y-10);

				return -mdist;
			}
		};

		DeterministicPlanner planner = new AStar(domain, goalCondition, hashingFactory, mdistHeuristic);
		Policy p = planner.planFromState(initialState);

		PolicyUtils.rollout(p, initialState, domain.getModel()).write(outputPath + "astar");

	}
 

/**
 * Initializes.
 * @param domain the domain in which to plan
 * @param gc should evaluate to true for goal states; false otherwise
 * @param hashingFactory the state hashing factory to use
 * @param heuristic the planning heuristic. Should return non-positive values.
 */
public IDAStar(SADomain domain, StateConditionTest gc, HashableStateFactory hashingFactory, Heuristic heuristic){
	
	this.deterministicPlannerInit(domain, gc, hashingFactory);
	
	this.heuristic = heuristic;
	nodeComparator = new PrioritizedSearchNode.PSNComparator();
	
}
 

/**
 * Initializes
 * @param domain the domain in which to plan
 * @param gc should evaluate to true for goal states; false otherwise
 * @param hashingFactory the state hashing factory to use
 * @param heuristic the planning heuristic. Should return non-positive values.
 * @param epsilon parameter > 1 indicating greediness; the larger the value the more greedy.
 * @param expectedDepth the expected depth of the plan
 */
public DynamicWeightedAStar(SADomain domain, StateConditionTest gc, HashableStateFactory hashingFactory, Heuristic heuristic, double epsilon, int expectedDepth) {
	super(domain, gc, hashingFactory, heuristic);
	this.epsilon = epsilon;
	this.expectedDepth = expectedDepth;
}
 

/**
 * Initializes the valueFunction.
 * @param domain the domain in which to plan
 * @param rf the reward function that represents costs as negative reward
 * @param gc should evaluate to true for goal states; false otherwise
 * @param hashingFactory the state hashing factory to use
 * @param heuristic the planning heuristic. Should return non-positive values.
 * @param costWeight a fraction 0 <= w <= 1. When w = 0, the search is fully greedy. When w = 1, the search is optimal and equivalent to A*.
 */
public WeightedGreedy(SADomain domain, RewardFunction rf, StateConditionTest gc, HashableStateFactory hashingFactory, Heuristic heuristic, double costWeight) {
	super(domain, gc, hashingFactory, heuristic);
	this.costWeight = costWeight;
}
 

/**
 * Initializes. Returned solution will be at most \epsilon times the optimal solution cost.
 * @param domain the domain in which to plan
 * @param gc should evaluate to true for goal states; false otherwise
 * @param hashingFactory the state hashing factory to use
 * @param heuristic the planning heuristic. Should return non-positive values.
 * @param epsilon parameter > 1. The larger the value the more greedy.
 */
public StaticWeightedAStar(SADomain domain, StateConditionTest gc, HashableStateFactory hashingFactory, Heuristic heuristic, double epsilon) {
	super(domain, gc, hashingFactory, heuristic);
	this.epsilonP1 = 1. + epsilon;
}
 

/**
 * Initializes A*. Goal states are indicated by gc evaluating to true. The costs are stored as negative rewards in the reward function.
 * By default there are no terminal states except teh goal states, so a terminal function is not taken.
 * @param domain the domain in which to plan
 * @param gc should evaluate to true for goal states; false otherwise
 * @param hashingFactory the state hashing factory to use
 * @param heuristic the planning heuristic. Should return non-positive values.
 */
public AStar(SADomain domain, StateConditionTest gc, HashableStateFactory hashingFactory, Heuristic heuristic){
	
	this.deterministicPlannerInit(domain, gc, hashingFactory);
	
	this.heuristic = heuristic;
	
}