type=float,
default=0.9,
help='Reward decay')
args = parse.parse_args()
# env setup
env = gym.make('MountainCar-v0')
env = env.unwrapped
# algorithm setup
method = args.method
if method == 'QL':
print("Use Q-Learning...")
print('--------------------------------')
RL = QLearningTable(actions=list(range(env.action_space.n)),
learning_rate=args.learning_rate,
reward_decay=args.reward_decay)
elif method == 'SARSA':
print("Use SARSA...")
print('--------------------------------')
RL = SarsaTable(actions=list(range(env.action_space.n)),
learning_rate=args.learning_rate,
reward_decay=args.reward_decay)
elif method == 'DQN':
print("Use DQN...")
print('--------------------------------')
env_shape = 0 if isinstance(
env.action_space.sample(),
int) else env.action_space.sample().shape # to confirm the shape
RL = DQN(action_n=env.action_space.n,
state_n=env.observation_space.shape[0],
def __init__(self, file_name, base_port=5006):
atexit.register(self.close)
self.port = base_port
self._buffer_size = 10240
self._loaded = False
self._open_socket = False
logger.info("unity env try created, socket with port:{}".format(
str(self.port)))
try:
# Establish communication socket
self._socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
self._socket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
self._socket.bind(("localhost", self.port))
self._open_socket = True
except socket.error:
self._open_socket = True
self.close()
raise socket.error(
"Couldn't launch new environment "
"You may need to manually close a previously opened environment "
"or use a different worker number.")
cwd = os.getcwd()
file_name = (file_name.strip().replace('.app', '').replace(
'.exe', '').replace('.x86_64', '').replace('.x86', ''))
true_filename = os.path.basename(os.path.normpath(file_name))
launch_string = None
if platform == "linux" or platform == "linux2":
candidates = glob.glob(os.path.join(cwd, file_name) + '.x86_64')
if len(candidates) == 0:
candidates = glob.glob(os.path.join(cwd, file_name) + '.x86')
if len(candidates) == 0:
candidates = glob.glob(file_name + '.x86_64')
if len(candidates) == 0:
candidates = glob.glob(file_name + '.x86')
if len(candidates) > 0:
launch_string = candidates[0]
elif platform == 'darwin':
candidates = glob.glob(
os.path.join(cwd, file_name + '.app', 'Contents', 'MacOS',
true_filename))
if len(candidates) == 0:
candidates = glob.glob(
os.path.join(file_name + '.app', 'Contents', 'MacOS',
true_filename))
if len(candidates) == 0:
candidates = glob.glob(
os.path.join(cwd, file_name + '.app', 'Contents', 'MacOS',
'*'))
if len(candidates) == 0:
candidates = glob.glob(
os.path.join(file_name + '.app', 'Contents', 'MacOS', '*'))
if len(candidates) > 0:
launch_string = candidates[0]
elif platform == 'win32':
candidates = glob.glob(os.path.join(cwd, file_name + '.exe'))
if len(candidates) == 0:
candidates = glob.glob(file_name + '.exe')
if len(candidates) > 0:
launch_string = candidates[0]
if launch_string is None:
self.close()
raise UnityEnvironmentException(
"Couldn't launch the {0} environment. "
"Provided filename does not match any environments.".format(
true_filename))
else:
# Launch Unity environment
proc1 = subprocess.Popen([launch_string, '--port', str(self.port)])
self._socket.settimeout(60)
try:
try:
self._socket.listen(1)
self._conn, _ = self._socket.accept()
self._conn.settimeout(30)
p = self._conn.recv(self._buffer_size).decode('utf-8')
p = json.loads(p)
# print p
except socket.timeout as e:
raise UnityTimeOutException(
"The Unity environment took too long to respond. Make sure {} does not need user interaction to "
"launch and that the Academy and the external Brain(s) are attached to objects in the Scene."
.format(str(file_name)))
self._data = {}
self._global_done = None
self._log_path = p["logPath"]
self._alpha = p["alpha"]
self._epsilon = p["epsilon"]
self._gamma = p["gamma"]
self._states = p["states"]
# self._num_states = len(self._states)
# for i in range(self._num_states):
# print "i:{0}, state:{1}".format(i,self._states[i])
self._actions = p["actions"]
self._brain = QLearningTable(self._actions, self._states,
self._alpha, self._gamma,
self._epsilon)
self._loaded = True
self._recv_bytes()
logger.info("started successfully!")
except UnityEnvironmentException:
proc1.kill()
self.close()
raise
parser.add_argument(
'--method',
default='DQN',
help=
'input method used to solve problem: Q-learning / Sarsa / SarsaLambda / DQN / Policy Gradient'
)
parser.add_argument('--episode',
default='5000',
help='input how many episodes to execute')
parser.add_argument('--test',
default='False',
help='is testing mode or not')
args = parser.parse_args()
if args.method == 'Q-learning':
RL = QLearningTable(range(0, env.action_space.n))
elif args.method == 'Sarsa':
RL = SarsaTable(range(0, env.action_space.n))
elif args.method == 'SarsaLambda':
RL = SarsaLambdaTable(range(0, env.action_space.n))
elif args.method == 'DQN':
if args.test == 'True':
RL = DeepQNetwork(env.action_space.n,
2,
lr=0.1,
batch_size=128,
reward_decay=0.9,
e_greedy=0.9,
replace_target_iter=300,
memory_size=3000,
e_greedy_increment=0.0001,
action = RL.choose_action(str(observation))
observation_, reward, done = game.ML_move(action)
if (reward > max):
max = reward
#print(str(observation) + " R" + str(reward))
RL.learn(str(observation), action, reward, str(observation_))
# swap observation
observation = observation_
# break while loop when end of this episode
if done:
print(max)
break
if __name__ == "__main__":
game = Game(288, 512)
game.setup()
reward = 0
done = False
RL = QLearningTable(actions=list(range(game.n_actions)))
#arcade.run()
#start_new_thread(main,())
thread1 = myThread()
#thread2 = myThread2()
thread1.start()
#thread2.start()
arcade.run()
for episode in range(100):
# initial observation
s_curr = maze._reset_maze()
while True:
maze._render()
# Get next action from the Q-table
action = rl.select_next_action(str(s_curr))
# take the action and observe the next state and reward
s_next, reward, isDone = maze._update_maze(action)
# learn from the feedback
rl.learn(str(s_curr), action, reward, str(s_next))
s_curr = s_next
if isDone:
break
print("Game over")
maze.destroy()
if __name__ == "__main__":
maze = Maze()
rl = QLearningTable(actions=list(range(len(maze.action_space))))
maze.after(100, update)
maze.mainloop()
step = 0
RL.learning_rate = lr[episode]
while True:
env.render() #env.update()
action = RL.choose_action(str(observation))
observation_, reward, done = env.step(action)
sum_reward += reward
step += 1
RL.learn(str(observation), action, reward, str(observation_), done)
observation = observation_
if done:
REWARD.append(sum_reward)
STEP.append(step)
break
REWARD = []
STEP = []
lr = [0.1] * 200 + [0.03] * 400 + [0.005] * 400 + [
0.0005
] * 800 #+[0.0001]*400#+[0.00001]*400
env = Maze(n_goals=2, n_agents=2,
origin_random=True) #random needs more training
RL = QLearningTable(actions=list(range(env.n_actions)))
env.after(100, update(len(lr)))
env.mainloop()
joblib.dump(RL, 'table2.pkl')
plt.plot(range(1, len(REWARD) + 1), REWARD)
plt.show()