class MobileAvoidance(EnvSpace):
def env_init(self):
self.env = CarEnv()
self.state = self.env.reset()
self.send_state_get_action(self.state)
self.var = 1
def on_predict_response(self, action):
self.var = self.var * 0.9995 if self.ep_use_step > cfg['DDPG'][
'memory_capacity'] else self.var
a = np.clip(np.random.normal(action, self.var), *self.env.action_bound)
next_state, reward, done, _ = self.env.step(action)
# print(next_state)
done = True if self.ep_use_step >= EP_MAXSTEP else done
self.send_train_get_action(self.state, action, reward, done,
next_state)
self.state = next_state
# print('self.env_name=',self.env_name)
if self.ep >= 30 and RENDER:
self.env.render()
if done:
self.state = self.env.reset()
self.send_state_get_action(self.state)
UPDATE_EVENT.clear() # no update now
ROLLING_EVENT.set() # start to roll out
workers = [Worker(wid=i) for i in range(N_WORKER)]
GLOBAL_UPDATE_COUNTER, GLOBAL_EP = 0, 0
GLOBAL_RUNNING_R = []
COORD = tf.train.Coordinator()
QUEUE = queue.Queue()
threads = []
for worker in workers: # worker threads
t = threading.Thread(target=worker.work, args=())
t.start()
threads.append(t)
# add a PPO updating thread
threads.append(threading.Thread(target=GLOBAL_PPO.update, ))
threads[-1].start()
COORD.join(threads)
# plot reward change and testing
plt.plot(np.arange(len(GLOBAL_RUNNING_R)), GLOBAL_RUNNING_R)
plt.xlabel('Episode')
plt.ylabel('Moving reward')
plt.ion()
plt.show()
while True:
s = env.reset()
for t in range(400):
env.render()
s, _, done = env.step(GLOBAL_PPO.choose_action(s))[0]
if done == 1:
break