为了账号安全,请及时绑定邮箱和手机立即绑定

CartPole-v0 的意外观察空间

CartPole-v0 的意外观察空间

慕斯709654 2022-01-05 10:47:48
我对通过自省获得的观察空间感到惊讶CartPole-v0。根据官方文档,这是我应该得到的:但是,这是我得到的:print(env.observation_space.low)print(env.observation_space.high)#[-4.8000002e+00 -3.4028235e+38 -4.1887903e-01 -3.4028235e+38]#[4.8000002e+00 3.4028235e+38 4.1887903e-01 3.4028235e+38]我正在使用最新版本的gym:!pip list|grep gymgym                 0.12.1   知道发生了什么吗?
查看完整描述

2 回答

?
莫回无

TA贡献1865条经验 获得超7个赞

如代码中所述,您似乎正在获得预期的行为,这有点令人困惑。一方面,对于 的观察空间是 [-4.8, 4.8] cart position,然而,实际上,当推车到达极限 [-2.4, 2.4] 时,情节应该结束。与pole angle情况类似。


class CartPoleEnv(gym.Env):

"""

Description:

    A pole is attached by an un-actuated joint to a cart, which moves along a frictionless track. The pendulum starts upright, and the goal is to prevent it from falling over by increasing and reducing the cart's velocity.


Source:

    This environment corresponds to the version of the cart-pole problem described by Barto, Sutton, and Anderson


Observation: 

    Type: Box(4)

    Num Observation                 Min         Max

    0   Cart Position             -4.8            4.8

    1   Cart Velocity             -Inf            Inf

    2   Pole Angle                 -24 deg        24 deg

    3   Pole Velocity At Tip      -Inf            Inf


Actions:

    Type: Discrete(2)

    Num Action

    0   Push cart to the left

    1   Push cart to the right


    Note: The amount the velocity that is reduced or increased is not fixed; it depends on the angle the pole is pointing. This is because the center of gravity of the pole increases the amount of energy needed to move the cart underneath it


Reward:

    Reward is 1 for every step taken, including the termination step


Starting State:

    All observations are assigned a uniform random value in [-0.05..0.05]


Episode Termination:

    Pole Angle is more than 12 degrees

    Cart Position is more than 2.4 (center of the cart reaches the edge of the display)

    Episode length is greater than 200

    Solved Requirements

    Considered solved when the average reward is greater than or equal to 195.0 over 100 consecutive trials.

"""

在此链接中,您可以阅读相关的 Github 问题。


*请注意,24 度等于 4.1887903e-01 弧度。


查看完整回答
反对 回复 2022-01-05
?
摇曳的蔷薇

TA贡献1793条经验 获得超6个赞

看起来像过时的文档,已经创建了一个问题:https : //github.com/openai/gym/issues/368


查看完整回答
反对 回复 2022-01-05
  • 2 回答
  • 0 关注
  • 413 浏览
慕课专栏
更多

添加回答

举报

0/150
提交
取消
意见反馈 帮助中心 APP下载
官方微信