package ch.idsia.ai.agents.ai;
import ch.idsia.ai.Evolvable;
import ch.idsia.ai.MLP;
import ch.idsia.ai.agents.Agent;
import ch.idsia.mario.environments.Environment;
/**
* Created by IntelliJ IDEA.
* User: julian
* Date: Apr 28, 2009
* Time: 2:09:42 PM
*/
public class SimpleMLPAgent implements Agent, Evolvable {
private MLP mlp;
private String name = "SimpleMLPAgent";
final int numberOfOutputs = 6;
final int numberOfInputs = 10;
public SimpleMLPAgent () {
mlp = new MLP (numberOfInputs, 10, numberOfOutputs);
}
private SimpleMLPAgent (MLP mlp) {
this.mlp = mlp;
}
public Evolvable getNewInstance() {
return new SimpleMLPAgent(mlp.getNewInstance());
}
public Evolvable copy() {
return new SimpleMLPAgent (mlp.copy ());
}
public void reset() {
mlp.reset ();
}
public void mutate() {
mlp.mutate ();
}
public boolean[] getAction(Environment observation) {
byte[][] scene = observation.getLevelSceneObservation(/*1*/);
double[] inputs = new double[]{probe(-1, -1, scene), probe(0, -1, scene), probe(1, -1, scene),
probe(-1, 0, scene), probe(0, 0, scene), probe(1, 0, scene),
probe(-1, 1, scene), probe(0, 1, scene), probe(1, 1, scene),
1};
double[] outputs = mlp.propagate (inputs);
boolean[] action = new boolean[numberOfOutputs];
for (int i = 0; i < action.length; i++) {
action[i] = outputs[i] > 0;
}
return action;
}
public AGENT_TYPE getType() {
return AGENT_TYPE.AI;
}
public String getName() {
return name;
}
public void setName(String name) {
this.name = name;
}
private double probe (int x, int y, byte[][] scene) {
int realX = x + 11;
int realY = y + 11;
return (scene[realX][realY] != 0) ? 1 : 0;
}
}