public static StatisticalSummary test(Agent controller, EvaluationOptions options, int seed) {
StatisticalSummary ss = new StatisticalSummary();
int kills = 0;
int timeLeft = 0;
int marioMode = 0;
float marioStatus = 0;
for (int i = 0; i < numberOfTrials; i++) {
options.setLevelLength(200 + (i * 128) + (seed % (i + 1)));
options.setLevelType(i % 3);
options.setLevelRandSeed(seed + i);
controller.reset();
options.setAgent(controller);
Evaluator evaluator = new Evaluator(options);
EvaluationInfo result = evaluator.evaluate().get(0);
ss.add(result.computeDistancePassed());
kills += result.computeKillsTotal();
timeLeft += result.timeLeft;
marioMode += result.marioMode;
marioStatus += result.marioStatus;
}
killsSum += kills;
marioStatusSum += marioStatus;
timeLeftSum += timeLeft;
marioModeSum += marioMode;
return ss;
}
public static void score(Agent agent, int startingSeed) {
killsSum = 0;
marioStatusSum = 0;
timeLeftSum = 0;
marioModeSum = 0;
TimingAgent controller = new TimingAgent(agent);
// RegisterableAgent.registerAgent (controller);
EvaluationOptions options = new CmdLineOptions(new String[0]);
// options.setNumberOfTrials(1);
options.setVisualization(false);
options.setFPS(GlobalOptions.MaxFPS);
System.out.println("Scoring controller " + controller + " with starting seed " + startingSeed);
double competitionScore = 0;
competitionScore += testConfig(controller, options, startingSeed, 0, false);
competitionScore += testConfig(controller, options, startingSeed, 3, false);
competitionScore += testConfig(controller, options, startingSeed, 5, false);
competitionScore += testConfig(controller, options, startingSeed, 10, false);
System.out.println("Competition score: " + competitionScore + "\n\n");
System.out.println("Number of levels cleared = " + marioStatusSum);
System.out.println("Additional (tie-breaker) info: ");
System.out.println("Total time left = " + timeLeftSum);
System.out.println("Total kills = " + killsSum);
System.out.println("Mario mode (small, large, fire) sum = " + marioModeSum);
}
public static double testConfig(
TimingAgent controller, EvaluationOptions options, int seed, int level, boolean paused) {
options.setLevelDifficulty(level);
options.setPauseWorld(paused);
StatisticalSummary ss = test(controller, options, seed);
double averageTimeTaken = controller.averageTimeTaken();
System.out.printf(
"Difficulty %d score %.4f (avg time %.4f)\n", level, ss.mean(), averageTimeTaken);
if (averageTimeTaken > 40) {
System.out.println(
"Maximum allowed average time is 40 ms per time step.\n" + "Controller disqualified");
System.exit(0);
}
return ss.mean();
}
public static void main(String[] args) {
EvaluationOptions options = new CmdLineOptions(new String[0]);
options.setNumberOfTrials(1);
options.setPauseWorld(false);
Evolvable initial = new SimpleMLPAgent();
// RegisterableAgent.registerAgent ((Agent) initial);
options.setMaxFPS(true);
options.setLevelDifficulty(0);
options.setVisualization(false);
ProgressTask task = new ProgressTask(options);
options.setLevelRandSeed((int) (Math.random() * Integer.MAX_VALUE));
ES es = new ES(task, initial, populationSize);
System.out.println("Evolving " + initial + " with task " + task);
final String fileName = "evolved" + (int) (Math.random() * Integer.MAX_VALUE) + ".xml";
for (int gen = 0; gen < generations; gen++) {
es.nextGeneration();
double bestResult = es.getBestFitnesses()[0];
System.out.println("Generation " + gen + " best " + bestResult);
Easy.save(es.getBests()[0], fileName);
}
Stats.main(new String[] {fileName, "1"});
}