def getFeatures(self, state, action): # extract the grid of food and wall locations and get the ghost locations food = state.getFood() walls = state.getWalls() ghosts = state.getGhostPositions() features = CustomCounter() features["bias"] = 1.0 # compute the location of pacman after he takes the action x, y = state.getPacmanPosition() dx, dy = Actions.directionToVector(action) next_x, next_y = int(x + dx), int(y + dy) # count the number of ghosts 1-step away features["#-of-ghosts-1-step-away"] = sum( (next_x, next_y) in Actions.getLegalNeighbors(g, walls) for g in ghosts) # if there is no danger of ghosts then add the food feature if not features["#-of-ghosts-1-step-away"] and food[next_x][next_y]: features["eats-food"] = 1.0 dist = closestFood((next_x, next_y), food, walls) if dist is not None: # make the distance a number less than one otherwise the update # will diverge wildly features["closest-food"] = float(dist) / (walls.width * walls.height) features.divideAll(10.0) return features
def __init__(self, epsilon=0.05, gamma=0.8, alpha=0.2, numTraining=900, extractor=SimpleExtractor(), **args): "You can initialize Q-values here..." args['epsilon'] = epsilon args['gamma'] = gamma args['alpha'] = alpha args['numTraining'] = numTraining self.featExtractor = extractor self.index = 0 # This is always Pacman self.weights = CustomCounter() self.q_values = CustomCounter() self.lastAction = None ReinforcementAgent.__init__(self, epsilon=epsilon, gamma=gamma, alpha=alpha, numTraining=numTraining) "*** YOUR CODE HERE ***"
def getFeatures(self, state, action): feats = CustomCounter() feats[state] = 1.0 feats['x=%d' % state[0]] = 1.0 feats['y=%d' % state[0]] = 1.0 feats['action=%s' % action] = 1.0 return feats
def __init__(self, **args): "You can initialize Q-values here..." ReinforcementAgent.__init__(self, **args) self.last_move = None self.last_index = -1 self.q_values = CustomCounter() self.best_state_action = None
def __init__(self, extractor='IdentityExtractor', **args): self.featExtractor = MyExtractor() # self.featExtractor = lookup(extractor, globals())() PacmanQAgent.__init__(self, **args) self.train = True self.start_with_trained_weights = True if self.train: self.weights = CustomCounter() if self.start_with_trained_weights: self.weights = {'eats-food': 3.744196232507296, 'closest-food': -0.04577473069235167, '#-of-ghosts-1-step-away': -10.027445395331362, 'bias': -6.304746841317069}
def __init__(self, extractor='IdentityExtractor', train=False, optilio=False, weights_values=[0, 0, 0, 0], **args): self.featExtractor = lookup(extractor, globals())() PacmanQAgent.__init__(self, **args) self.weights = CustomCounter() self.weight = 0 self.optilio = optilio self.train = train if self.train: for number, feature in zip(weights_values, self.weights): self.weights[feature] = float(number)
def __init__(self, extractor='IdentityExtractor', **args): self.featExtractor = MyExtractor() # self.featExtractor = lookup(extractor, globals())() PacmanQAgent.__init__(self, **args) self.train = False self.start_with_trained_weights = True if self.train: self.weights = CustomCounter() if self.start_with_trained_weights: # self.weights = {'capsule-is-nearby': 0.0, 'bias': -5.848186678365304, 'ghost-is-nearby': -47.767678591296, 'eats-food': 4.2052171620675125, '#-of-ghosts-1-step-away': -10.053403840432816, 'closest-food': -0.09139458036757105} self.weights = { 'ghost-is-nearby': -19.71752071573443, 'bias': -5.853012565325543, 'eats-food': 4.2186468885699115, '#-of-ghosts-1-step-away': -10.029721097702556, 'closest-food': -0.09146973122692789 }
def __init__(self, epsilon=0.25, gamma=0.8, alpha=0.2, numTraining=0, extractor=IdentityExtractor(), **args): "You can initialize Q-values here..." args['epsilon'] = epsilon args['gamma'] = gamma args['alpha'] = alpha args['numTraining'] = numTraining self.featExtractor = extractor self.index = 0 # This is always Pacman self.weights = CustomCounter() ReinforcementAgent.__init__(self, **args) "*** YOUR CODE HERE ***"
def __init__(self, extractor='IdentityExtractor', train=False, optilio=False, weights_values=[0, 0, 0, 0, 0], **args): self.featExtractor = lookup(extractor, globals())() PacmanQAgent.__init__(self, **args) self.weights = CustomCounter() self.weight = 0 self.optilio = optilio self.train = train if (not self.train ) or self.optilio or weights_values != [0, 0, 0, 0, 0]: for number, feature in zip(weights_values, [ "bias", "#-of-ghosts-1-step-away", "eats-food", "closest-food", "capsules" ]): self.weights[feature] = float(number) if not self.optilio: # print(self.weights) pass
def getFeatures(self, state, action): feats = CustomCounter() feats[(state, action)] = 1.0 return feats
def getFeatures(self, state, action): # extract the grid of food and wall locations and get the ghost locations food = state.getFood() walls = state.getWalls() ghosts = state.getGhostPositions() features = CustomCounter() features["bias"] = 1.0 # compute the location of pacman after he takes the action x, y = state.getPacmanPosition() dx, dy = Actions.directionToVector(action) next_x, next_y = int(x + dx), int(y + dy) not_scared_ghosts_positions = [ g_s.getPosition() for g_s in state.getGhostStates() if not g_s.scaredTimer ] # count the number of ghosts 1-step away # get position of ghost only when it is not scared # in this way pacman may learn to eat scared ghost features["#-of-ghosts-1-step-away"] = sum( (next_x, next_y) in Actions.getLegalNeighbors(g_s, walls) for g_s in not_scared_ghosts_positions) # ghost_quarters = [get_quarter_from_position(ghost_position, walls) # for ghost_position in state.getGhostPositions()] # # if all(ghost_quarters) == get_quarter_from_position((x, y), walls): # features['pacman-and-ghosts-in-the-same-region'] = 1.0 # else: # features['pacman-and-ghosts-in-the-same-region'] = 0.0 ghost_distance_limit = 3 nearest_ghost = distance_to_the_nearest_item_from_list( (next_x, next_y), not_scared_ghosts_positions, walls, cutout=3) if nearest_ghost is not None: nearest_ghost = max(nearest_ghost - 1, 0) if nearest_ghost < 1: features['ghost-is-nearby'] = 1 else: features['ghost-is-nearby'] = ( ghost_distance_limit - nearest_ghost) / ghost_distance_limit else: features['ghost-is-nearby'] = 0 # nearest_capsule = distance_to_the_nearest_item_from_list((next_x, next_y), state.getCapsules(), walls, cutout=0) ## set 1.0 if the nearest capsule is closer than the nearest ghost # if nearest_capsule is not None and nearest_ghost is None or \ # nearest_capsule is not None and nearest_ghost is not None and nearest_capsule < nearest_ghost: # features["capsule-is-nearby"] = 1.0 # features["capsule-is-nearby"] = 1.0 if nearest_capsule is not None else 0.0 # if there is no danger of ghosts then add the food feature if not features["#-of-ghosts-1-step-away"] and food[next_x][next_y]: features["eats-food"] = 1.0 else: features["eats-food"] = 0.0 dist = closestFood((next_x, next_y), food, walls) if dist is not None: # make the distance a number less than one otherwise the update # will diverge wildly features["closest-food"] = float(dist) / (walls.width * walls.height) features.divideAll(10.0) return features
def getFeatures(self, state, action): # extract the grid of food and wall locations and get the ghost locations food = state.getFood() walls = state.getWalls() ghosts = state.getGhostPositions() capsulesLeft = len(state.getCapsules()) scaredGhost = [] activeGhost = [] features = CustomCounter() for ghost in state.getGhostStates(): if not ghost.scaredTimer: activeGhost.append(ghost) else: #print (ghost.scaredTimer) scaredGhost.append(ghost) pos = state.getPacmanPosition() def getManhattanDistances(ghosts): return map(lambda g: manhattan_distance(pos, g.getPosition()), ghosts) distanceToClosestActiveGhost = distanceToClosestScaredGhost = 0 features["bias"] = 1.0 # compute the location of pacman after he takes the action x, y = state.getPacmanPosition() dx, dy = Actions.directionToVector(action) next_x, next_y = int(x + dx), int(y + dy) # count the number of ghosts 1-step away features["#-of-ghosts-1-step-away"] = sum( (next_x, next_y) in Actions.getLegalNeighbors(g, walls) for g in ghosts) # if there is no danger of ghosts then add the food feature if not features["#-of-ghosts-1-step-away"] and food[next_x][next_y]: features["eats-food"] = 1.0 dist = closestFood((next_x, next_y), food, walls) if dist is not None: # make the distance a number less than one otherwise the update # will diverge wildly features["closest-food"] = float(dist) / (walls.width * walls.height) if scaredGhost: # and not activeGhost: distanceToClosestScaredGhost = min( getManhattanDistances(scaredGhost)) if activeGhost: distanceToClosestActiveGhost = min( getManhattanDistances(activeGhost)) else: distanceToClosestActiveGhost = 10 features["capsules"] = capsulesLeft #features["dist-to-closest-active-ghost"] = 2*(1./distanceToClosestActiveGhost) if distanceToClosestScaredGhost <= 8 and distanceToClosestActiveGhost >= 2: #features["#-of-ghosts-1-step-away"] >= 1: features["#-of-ghosts-1-step-away"] = 0 features["eats-food"] = 0.0 #features["closest-food"] = 0 features.divideAll(10.0) return features
def __init__(self, extractor='IdentityExtractor', **args): self.featExtractor = lookup(extractor, globals())() PacmanQAgent.__init__(self, **args) self.weights = CustomCounter()