pong_wrapper.py 文件源码

def step(self, action):
    screens = []
    total_reward = 0

    for t in range(4):
      screen = self.get_screen()
      screens.append(screen)
      _, reward, done, info = self.env.step(self.action_mapping[action])
      total_reward += reward
      if done or total_reward:
        if done:
          self.env.reset()
        for _ in range(20):
          self.env.step(0)
        for _ in range(3 - t):
          screens.append(screen)
        break

    screens = np.asarray(screens).astype(np.float)
    return screens, total_reward, done, info