OpenAI健身房自定义环境:具有实值的离散观察空间

import gym
from gym import spaces
import numpy as np

class CustomEnv(gym.Env):
    def __init__(self):
        super(CustomEnv, self).__init__()
        # 定义离散观察空间和动作空间
        self.observation_space = spaces.Discrete(5)  # 假设有5个离散状态
        self.action_space = spaces.Discrete(2)      # 假设有2个离散动作
        
        self.state = None
    
    def reset(self):
        self.state = np.random.choice(self.observation_space.n)  # 随机初始化状态
        return self.state
    
    def step(self, action):
        assert self.action_space.contains(action), "%r (%s) invalid" % (action, type(action))
        
        # 根据动作更新状态（这里仅为示例，实际情况可能更复杂）
        if action == 0:
            self.state = max(self.state - 1, 0)
        else:
            self.state = min(self.state + 1, self.observation_space.n - 1)
        
        # 定义奖励函数（这里仅为示例）
        reward = 1 if self.state == 4 else 0
        
        done = self.state == 4  # 假设到达状态4时任务完成
        
        return self.state, reward, done, {}
    
    def render(self, mode='human'):
        print(f"Current state: {self.state}")

# 使用自定义环境
env = CustomEnv()
observation = env.reset()
for _ in range(10):
    env.render()
    action = env.action_space.sample()  # 随机选择动作
    observation, reward, done, info = env.step(action)
    if done:
        observation = env.reset()
env.close()