model_based_learning.py 文件源码

def main():
  env = gym.make('Stochastic-4x4-FrozenLake-v0')
  policy = learn_with_mdp_model(env)
  render_single(env, policy)

  # for i in range(10):
  #   print('\n%d' % i)
  #   env.render()
  #   print(env.step(env.action_space.sample()))
  # env.render()
  # for init_state in env.P.keys():
  #   for action in env.P[init_state]:
  #     print("\nState: %d, action: %d" % (init_state, action))
  #     for next_state in env.P[init_state][action]:
  #       print(next_state)
  # for _ in range(10):
  #   env.render()
  #   env.step(env.action_space.sample())