public Action policy(State observation, float reward) {
Map<Float, Action> options = new HashMap<Float, Action>();
scala.collection.Iterator<Action> it = actions().iterator();
while (it.hasNext()) {
Action action = it.next();
options.put(expectedReward(observation, action), action);
}
Action bestGuess = options.get(new TreeSet<Float>(options.keySet()).last());
Action decision = bestGuess;
if (random.nextDouble() < randomFactor) {
decision = randomAction();
}
Situation currentSA = new Situation(observation, decision);
float currentExpected = expectedReward(observation, decision);
float lastExpected = expectedReward(lastSA.getKey(), lastSA.getValue());
float newQValue = lastExpected + alpha * (reward + gamma * currentExpected - lastExpected);
qTable.put(lastSA, newQValue); // update the Q-Table
lastSA = currentSA;
return decision;
}
