python人工智能 RL OpenAI Gym 强化学习: 用 OpenAI Gym 实现一个 RL 强化学习 任务，代码方案分享1(图文详解)

import gym

class Maze(gym.Env):
    def __init__(self):
        # 初始化迷宫的大小和其他参数
        self.grid_size = 5
        self.start_position = (0, 0)
        self.goal_position = (4, 4)

        # 定义动作空间和观察空间
        self.action_space = gym.spaces.Discrete(4)  # 上下左右四个动作
        self.observation_space = gym.spaces.Discrete(self.grid_size ** 2)  # 格子的总数

        # 初始化智能体的位置
        self.agent_position = self.start_position

    def reset(self):
        # 重置智能体的位置到起始位置
        self.agent_position = self.start_position
        return self._get_observation()

    def step(self, action):
        # 执行动作，更新智能体的位置
        if action == 0:  # 上
            new_position = (self.agent_position[0], max(0, self.agent_position[1] - 1))
        elif action == 1:  # 下
            new_position = (self.agent_position[0], min(self.grid_size - 1, self.agent_position[1] + 1))
        elif action == 2:  # 左
            new_position = (max(0, self.agent_position[0] - 1), self.agent_position[1])
        elif action == 3:  # 右
            new_position = (min(self.grid_size - 1, self.agent_position[0] + 1), self.agent_position[1])

        self.agent_position = new_position

        # 判断是否到达目标位置
        done = (self.agent_position == self.goal_position)

        # 计算奖励
        if done:
            reward = 1.0
        else:
            reward = -0.1

        return self._get_observation(), reward, done, {}

    def _get_observation(self):
        # 将智能体的位置转换为观察
        return self.agent_position[0] * self.grid_size + self.agent_position[1]

    def render(self, mode='human'):
        # 在终端打印迷宫的状态
        for i in range(self.grid_size):
            for j in range(self.grid_size):
                if (i, j) == self.agent_position:
                    print('A', end=' ')
                elif (i, j) == self.goal_position:
                    print('G', end=' ')
                else:
                    print('.', end=' ')
            print()

import gym

env = Maze()

# 训练智能体的代码
for episode in range(10):
    observation = env.reset()
    done = False
    steps = 0

    while not done:
        action = env.action_space.sample()  # 随机选择动作
        observation, reward, done, _ = env.step(action)
        steps += 1

        env.render()  # 打印迷宫状态

    print(f"Episode {episode + 1} finished in {steps} steps.\n")