RL: my_random_agent.py
作者: 魏鹏飞 | 来源:发表于2020-04-08 20:44 被阅读0次
RL: my_random_agent.py
mac 本机mysql无法启动
强化学习
RL
RL
RL
rl
10.31 背
Arrow Of RL
再见RL
Keywords:

Agent、action_space、sample、Monitor、episode、maxstep、observation、reward、done、info、step、
my_random_agent.py

import gym
import argparse
## random agent
class RandomAgent(object):
    def __init__(self, action_space):
        self.action_space = action_space

    def act(self, observation, reward, done):
        return self.action_space.sample()

class BiasedAgent(object):
    def __init__(self, action_space):
        self.action_space = action_space
        self.action_always = self.action_space.sample()
    def act(self, observation, reward, done):
        return self.action_always

if __name__ == '__main__':

    parser = argparse.ArgumentParser()
    parser.add_argument('--display', action='store_true')
    parser.add_argument('game', nargs="?", default="CartPole-v0")
    args = parser.parse_args()

    env = gym.make(args.game)
    # env = gym.wrappers.Monitor(env, 'video', force = True)
    num_episodes = 20
    num_maxstep = 100

    agent_id = 1
    if agent_id == 1:
        agent = RandomAgent(env.action_space)
    elif agent_id == 2:
        agent = BiasedAgent(env.action_space)

    reward = 0
    done = False

    for i_episode in range(num_episodes):
        observation = env.reset()
        for t in range(num_maxstep):
            env.render()
            action = agent.act(observation, reward, done)
            observation, reward, done, info = env.step(action)
            print('episode {}-step {}, taking action {}, observation {}'.format(i_episode, t, action, observation))
            if done:
                print("Episode finished after {} timesteps".format(t+1))
                break
    env.close()


# Results:
python my_random_agent.py CartPole-v0

episode 0-step 0, taking action 0, observation [ 0.02091297 -0.18428767 -0.03879215  0.27603339]
episode 0-step 1, taking action 0, observation [ 0.01722722 -0.37883531 -0.03327148  0.55623328]
episode 0-step 2, taking action 1, observation [ 0.00965051 -0.18326243 -0.02214682  0.2532563 ]
episode 0-step 3, taking action 0, observation [ 0.00598526 -0.37806126 -0.01708169  0.53887227]
episode 0-step 4, taking action 0, observation [-0.00157596 -0.57293896 -0.00630424  0.8261244 ]
episode 0-step 5, taking action 0, observation [-0.01303474 -0.76797413  0.01021824  1.11681791]
episode 0-step 6, taking action 0, observation [-0.02839422 -0.96322869  0.0325546   1.41268857]
episode 0-step 7, taking action 0, observation [-0.0476588  -1.15873867  0.06080837  1.7153675 ]
episode 0-step 8, taking action 1, observation [-0.07083357 -0.96436472  0.09511572  1.44221211]
episode 0-step 9, taking action 0, observation [-0.09012087 -1.16052037  0.12395997  1.76303751]
episode 0-step 10, taking action 0, observation [-0.11333127 -1.35680743  0.15922072  2.09155982]
episode 0-step 11, taking action 1, observation [-0.14046742 -1.16360922  0.20105191  1.85204148]
episode 0-step 12, taking action 0, observation [-0.16373961 -1.36029526  0.23809274  2.19983962]
Episode finished after 13 timesteps
episode 1-step 0, taking action 0, observation [ 0.00422353 -0.14742106 -0.01258727  0.26181792]
episode 1-step 1, taking action 0, observation [ 0.00127511 -0.34236109 -0.00735091  0.55050425]
episode 1-step 2, taking action 0, observation [-0.00557211 -0.53737902  0.00365918  0.8408621 ]
episode 1-step 3, taking action 0, observation [-0.01631969 -0.73255074  0.02047642  1.13469351]
episode 1-step 4, taking action 1, observation [-0.03097071 -0.53770262  0.04317029  0.84850225]
episode 1-step 5, taking action 1, observation [-0.04172476 -0.34319522  0.06014033  0.56970111]
episode 1-step 6, taking action 1, observation [-0.04858866 -0.148966    0.07153436  0.29655459]
episode 1-step 7, taking action 0, observation [-0.05156798 -0.34503095  0.07746545  0.61091302]
episode 1-step 8, taking action 0, observation [-0.0584686  -0.54114531  0.08968371  0.92695443]
episode 1-step 9, taking action 0, observation [-0.06929151 -0.73735632  0.1082228   1.24641981]
episode 1-step 10, taking action 0, observation [-0.08403864 -0.93368695  0.13315119  1.57094827]
episode 1-step 11, taking action 1, observation [-0.10271237 -0.74038174  0.16457016  1.32258791]
episode 1-step 12, taking action 1, observation [-0.11752001 -0.54767645  0.19102192  1.08560099]
episode 1-step 13, taking action 1, observation [-0.12847354 -0.35551642  0.21273394  0.85842335]
Episode finished after 14 timesteps
......
......
......
episode 18-step 0, taking action 0, observation [-0.037643   -0.22180154 -0.04034859  0.26337507]
episode 18-step 1, taking action 0, observation [-0.04207903 -0.41632503 -0.03508109  0.54306356]
episode 18-step 2, taking action 0, observation [-0.05040553 -0.61093685 -0.02421982  0.82448996]
episode 18-step 3, taking action 1, observation [-0.06262427 -0.41549214 -0.00773002  0.52428894]
episode 18-step 4, taking action 0, observation [-0.07093411 -0.61050445  0.00275576  0.81452607]
episode 18-step 5, taking action 1, observation [-0.0831442  -0.41542035  0.01904628  0.52271122]
episode 18-step 6, taking action 1, observation [-0.09145261 -0.22057158  0.02950051  0.23609034]
episode 18-step 7, taking action 1, observation [-0.09586404 -0.02588325  0.03422231 -0.04714319]
episode 18-step 8, taking action 1, observation [-0.09638171  0.16873169  0.03327945 -0.32883528]
episode 18-step 9, taking action 1, observation [-0.09300707  0.36336446  0.02670274 -0.61084042]
episode 18-step 10, taking action 1, observation [-0.08573978  0.5581032   0.01448594 -0.89499472]
episode 18-step 11, taking action 1, observation [-0.07457772  0.75302576 -0.00341396 -1.18308917]
episode 18-step 12, taking action 1, observation [-0.0595172   0.94819184 -0.02707574 -1.47684029]
episode 18-step 13, taking action 0, observation [-0.04055337  0.75341084 -0.05661255 -1.19273518]
episode 18-step 14, taking action 1, observation [-0.02548515  0.94921857 -0.08046725 -1.50261142]
episode 18-step 15, taking action 0, observation [-0.00650078  0.75516034 -0.11051948 -1.23609781]
episode 18-step 16, taking action 0, observation [ 0.00860243  0.56161839 -0.13524144 -0.97998272]
episode 18-step 17, taking action 0, observation [ 0.0198348   0.36854293 -0.15484109 -0.73265393]
episode 18-step 18, taking action 1, observation [ 0.02720565  0.56542679 -0.16949417 -1.06978805]
episode 18-step 19, taking action 1, observation [ 0.03851419  0.76233494 -0.19088993 -1.41051083]
episode 18-step 20, taking action 0, observation [ 0.05376089  0.57002289 -0.21910015 -1.18306398]
Episode finished after 21 timesteps
episode 19-step 0, taking action 0, observation [-0.00408516 -0.16827423  0.01038206  0.33411965]
episode 19-step 1, taking action 0, observation [-0.00745064 -0.36354239  0.01706445  0.63005838]
episode 19-step 2, taking action 0, observation [-0.01472149 -0.55889826  0.02966562  0.92806621]
episode 19-step 3, taking action 0, observation [-0.02589945 -0.7544079   0.04822695  1.22992205]
episode 19-step 4, taking action 1, observation [-0.04098761 -0.55993844  0.07282539  0.95273026]
episode 19-step 5, taking action 0, observation [-0.05218638 -0.75596076  0.09187999  1.26737611]
episode 19-step 6, taking action 1, observation [-0.0673056  -0.56212471  0.11722751  1.00482318]
episode 19-step 7, taking action 0, observation [-0.07854809 -0.75860065  0.13732398  1.33190038]
episode 19-step 8, taking action 1, observation [-0.0937201  -0.56545138  0.16396199  1.08515046]
episode 19-step 9, taking action 1, observation [-0.10502913 -0.37282697  0.185665    0.8480781 ]
episode 19-step 10, taking action 0, observation [-0.11248567 -0.56992986  0.20262656  1.19292367]
episode 19-step 11, taking action 0, observation [-0.12388427 -0.76701546  0.22648503  1.54166931]
Episode finished after 12 timesteps