custom_environments

# 为 Stable Baselines3 创建自定义环境 本指南提供了为 Stable Baselines3 创建兼容的自定义 Gymnasium 环境的全面信息。 ## 环境结构 ### 必需方法 每个自定义环境必须继承自 `gymnasium.Env` 并实现： python import gymnasium as gym from gymnasium import spaces import numpy as np class CustomEnv(gym.Env): def __init__(self): """初始化环境，定义 action_space 和 observation_space""" super().__init__() self.action_space = spaces.Discrete(4) self.observation_space = spaces.Box(low=0, high=1, shape=(4,), dtype=np.float32) def reset(self, seed=None, options=None): """将环境重置为初始状态""" super().reset(seed=seed) observation = self.observation_space.sample() info = {} return observation, info def step(self, action): """执行一个时间步""" observation = self.observation_space.sample() reward = 0.0 terminated = False # 片段自然结束 truncated = False # 片段因时间限制结束 info = {} return observation, reward, terminated, truncated, info def render(self): """可视化环境（可选）""" pass def close(self): """清理资源（可选）""" pass ### 方法详情 #### `__init__(self, ...)` **目的：** 初始化环境并定义空间。 **要求：** - 必须调用 `super().__init__()` - 必须定义 `self.action_space` - 必须定义 `self.observation_space` **示例：** python def __init__(self, grid_size=10, max_steps=100): super().__init__() self.grid_size = grid_size self.max_steps = max_steps self.current_step = 0 # 定义空间 self.action_space = spaces.Discrete(4) self.observation_space = spaces.Box( low=0, high=grid_size-1, shape=(2,), dtype=np.float32 ) #### `reset(self, seed=None, options=None)` **目的：** 将环境重置为初始状态。 **要求：** - 必须调用 `super().reset(seed=seed)` - 必须返回 `(observation, info)` 元组 - 观测值必须符合 `observation_space` - Info 必须是一个字典（可以是空的） **示例：** python def reset(self, seed=None, options=None): super().reset(seed=seed) # 初始化状态 self.agent_pos = self.np_random.integers(0, self.grid_size, size=2) self.go