action.java

Reinforcement Learning

public class Action
{
	final static int numActions = 4;
	final static int UP = 0, RIGHT = 1, DOWN = 2, LEFT = 3;
	
	public Action()
	{	}
	
	/**
	 * Performs the specified action (UP, RIGHT, DOWN, LEFT) on the state 'st' 
	 * and returns the resulting new state 
	 */
	public static State performAction(State st, int action)
	{
		State newSt;
		switch(action)
		{
			case UP:
				newSt = new State(st.x, st.y+1); 
				break;
			case RIGHT:
				newSt = new State(st.x+1, st.y);
				break;
			case DOWN:
				newSt = new State(st.x, st.y-1);
				break;
			case LEFT:
				newSt = new State(st.x-1, st.y);
				break;
			default:
				newSt = st;
		}
		return newSt;
	}
	
	public static State performAction(State st, int action, double pjog)
	{
	    //Random randomGen = new Random();
	    double rand = Math.random();
		State newSt;
		double randActProb=pjog/(Action.numActions);
		int choosenAction = action;
		
		for(int i=0;i<Action.numActions;i++) {
			if (rand < (i+1)*randActProb ) {
				choosenAction=i;
				break;
			}
		}
		Utility.show("==================");
		Utility.show("Action to be Taken:"+action);
		Utility.show("Rand"+rand);
		Utility.show("Action Taken:"+choosenAction);
		Utility.show("==================");
		
		newSt = performAction(st, choosenAction);
		return newSt;
	}

}