gamma=0.99,
minibatch_size=64,
max_episodes=50000,
max_episode_len=1000,
# Exploration Strategies
exploration_action_noise_type='ou_0.2',
exploration_epsilon_greedy_type='none',
# Save Summaries
save_dir='../ROM_Experiment_results/LASAgentActorCritic/',
experiment_runs='run3',
# Save and Restore Actor-Critic Model
restore_actor_model_flag=False,
restore_critic_model_flag=False)
# Step counter
i = 1
done = False
reward_for_LAS = 0
while not done:
# LAS interacts with environment.
actionLAS = LASAgent1.perceive_and_act(observation_For_LAS,
reward_for_LAS, done)
# delay the observing of consequence of LASAgent's action
observation_For_LAS, reward_for_LAS, done, info = envLAS.step(
actionLAS)
print("LAS Step: {}, reward: {}".format(i, reward_for_LAS))
i += 1
envLAS.destroy()
reward = 0
done = False
start_time = time.time()
try:
for i in range(max_episodes):
observation = env.reset()
ep_reward = 0
for j in range(max_episode_len):
if render_env == True:
env.render()
# Added exploration noise
action = LASAgent.perceive_and_act(observation, reward,
done)
observation, reward, done, info = env.step(action[0])
ep_reward += reward
if done or j == (max_episode_len - 1):
print('| Reward: {:d} | Episode: {:d} '.format(
int(ep_reward), i))
episod_reward_memory.append(ep_reward)
plot_cumulative_reward(episod_reward_memory)
break
#time.sleep(0.5)
print("Time elapsed:{}".format(time.time() - start_time))
except KeyboardInterrupt:
sess.close()
env.destroy()
print("Shut Down.")
Created on Wed May 9 08:52:36 2018
@author: jack.lingheng.meng
"""
from Environment.LASEnv import LASEnv
from LASAgent.RandomLASAgent import RandomLASAgent
if __name__ == '__main__':
# Instantiate environment object
envLAS = LASEnv('127.0.0.1', 19997)
observationForLAS = envLAS.reset()
# Iinstantiate LAS-agent
LASAgent1 = RandomLASAgent(envLAS.observation_space, envLAS.action_space)
# Step counter
i = 1
done = False
rewardLAS = 0
while not done:
actionLAS = LASAgent1.interact(observationForLAS, rewardLAS, done)
observationForLAS, rewardLAS, done, info = envLAS.step(actionLAS)
print("LAS Step: {}, reward: {}".format(i, rewardLAS))
i += 1
envLAS.destroy()