openaigym

import gym
import numpy as np
env = gym.make('Breakout-v0')
print('observation space:', env.observation_space)
print('action space:', env.action_space)

obs = env.reset()
# env.render()
print('initial observation:', obs)

action = env.action_space.sample()
obs, r, done, info = env.step(action)
print('next observation:', obs)
print('reward:', r)
print('done:', done)
print('info:', info)

# Start virtual display
from pyvirtualdisplay import Display
display = Display(visible=0, size=(1024, 768))
display.start()

frames = []
for i in range(3):
    obs = env.reset()
    done = False
    R = 0
    t = 0
    while not done and t < 200:
        frames.append(env.render(mode = 'rgb_array'))
        obs, r, done, _ = env.step(env.action_space.sample())
        R += r
        t += 1
    print('test episode:', i, 'R:', R)
 
env.render()

import matplotlib.pyplot as plt
import matplotlib.animation
import numpy as np
from IPython.display import HTML

plt.figure(figsize=(frames[0].shape[1] / 72.0, frames[0].shape[0] / 72.0), dpi = 72)
patch = plt.imshow(frames[0])
plt.axis('off')
animate = lambda i: patch.set_data(frames[i])
ani = matplotlib.animation.FuncAnimation(plt.gcf(), animate, frames=len(frames), interval = 50)
HTML(ani.to_jshtml())

import gym
from IPython import display
import matplotlib.pyplot as plt
%matplotlib inline

env = gym.make('CartPole-v0')
env.reset()
for _ in range(50):
    plt.imshow(env.render(mode='rgb_array'))
    display.clear_output(wait=True)
    display.display(plt.gcf())
    env.step(env.action_space.sample())
    plt.close()

動画保存

import numpy as np
import gym

def sigmoid(x):
    return 1.0 / (1.0 + np.exp(-x))

env = gym.make('CartPole-v1')
desired_state = np.array([0, 0, 0, 0])
desired_mask = np.array([0, 0, 1, 0])

P, I, D = 0.1, 0.01, 0.5
env = gym.wrappers.Monitor(env, 'video/',video_callable=lambda episode_id: True,force = True)
for i_episode in range(20):
    state = env.reset()
    integral = 0
    derivative = 0
    prev_error = 0
    for t in range(500):
        env.render()
        error = state - desired_state

        integral += error
        derivative = error - prev_error
        prev_error = error

        pid = np.dot(P * error + I * integral + D * derivative, desired_mask)
        action = sigmoid(pid)
        action = np.round(action).astype(np.int32)

        state, reward, done, info = env.step(action)
        if done:
            print("Episode finished after {} timesteps".format(t+1))
            break

env.close()

動画保存

# https://github.com/TTitcombe/docker_openai_gym
import gym
import torch

env_to_wrap = gym.make("BankHeist-ram-v0")
env = gym.wrappers.Monitor(env_to_wrap, "someDir2")
frame = env.reset()
is_done = False
while not is_done:
  action = env.action_space.sample()
  _, _, is_done, _ = env.step(action)
env.close()
env_to_wrap.close()

https://gist.github.com/HenryJia/23db12d61546054aa43f8dc587d9dc2c
http://korechipostit.hatenablog.com/entry/2018/11/05/190000
https://colab.research.google.com/drive/1GLlB53gvZaUyqMYv8GmZQJmshRUzV_tg#scrollTo=ZAcCvXYJlJ8I