Gym v0.26.0自定义迷宫环境reset()报TypeError:意外'seed'参数

我基于OpenAI Gym v0.26.0开发了自定义迷宫环境MazeEnv，测试时触发TypeError，提示reset() got an unexpected keyword argument 'seed'。我已经按照Gym官方文档编写代码，以下是环境类代码、测试代码和完整报错栈：

自定义环境类代码 #

import gym
from gym import spaces
import pygame
import numpy as np
import math
import random

class MazeEnv(gym.Env):
    metadata = {"render_modes": ["human", "rgb_array"], "render_fps": 4}
    
    
    def __init__(self, render_mode=None, size=10.0):
        self.size = size
        self.window_size = 512
        
        self.observation_space = spaces.Dict(
            {
                "agent": spaces.Box(low=0.0, high=10.0, shape=(2,), dtype=np.float32),
                "target": spaces.Box(low=0.0, high=10.0, shape=(2,), dtype=np.float32)
                }
            )
        
        self.action_space = spaces.Box(low=-math.pi, high=math.pi, shape = (1,))
        
        assert render_mode is None or render_mode in self.metadata["render_modes"]
        self.render_mode = render_mode
        
        self.window = None
        self.clock = None
        
        
    def _get_obs(self):
        return {"agent": self._agent_location, "target": self._target_location}
    
    
    def _get_info(self):
        return {"distance": np.linalg.norm(self._agent_location - self._target_location, ord=2)}
    
    
    def reset(self, seed=None, options=None):
        super().reset(seed=seed)
        
        # agents starting location
        self._agent_location = [0.0, 10.0]
        self._target_location = [10.0, 0.0]
        
        observation = self._get_obs()
        info = self._get_info()
        
        if self.render_mode == "human":
            self._render_frame()
            
        return observation, info
    
    
    def step(self, action):
        terminated = False
        
        angle = random.gauss(action, math.pi/6)
        direction = [math.cos(angle)/2, math.sin(angle)/2]
        
        if self._agent_location[0] + direction[0] < 0.0 or self._agent_location[1] + direction[1] < 0.0 or self._agent_location[0] + direction[0] > 10.0 or self._agent_location[1] + direction[1] > 10.0:
            self._agent_location = [0.0, 10.0] ######### POSSIBLY RESET HERE?
        else:
            self._agent_location = self._agent_location + direction
            
        observation = self._get_obs()
        info = self._get_info()
        
        if info["distance"] < 0.5:
            terminated = True
        reward = 1 if terminated else 0
        
        if self.render_mode == "human":
            self._render_frame()
            
        return observation, reward, terminated, False, info
    
    
    def render(self):
        if self.render_mode == "rgb_array":
            return self._render_frame()
        
    def _render_frame(self):
        if self.window is None and self.render_mode == "human":
            pygame.init()
            pygame.display.init()
            self.window = pygame.display.set_mode((self.window_size, self.window_size))
        if self.clock is None and self.render_mode == "human":
            self.clock = pygame.time.Clock()
            
        canvas = pygame.Surface((self.window_size, self.window_size))
        canvas.fill((255, 255, 255))
        
        pygame.draw.circle(canvas, (255, 0, 0), self._target_location, 0.5, draw_top_left=True)
        
        pygame.draw.circle(canvas, (0,0,255), self._agent_location, 0.02)
        
        pygame.draw.line(canvas, (0,0,0), (0.0,0.0), (0.0,10.0))
        pygame.draw.line(canvas, (0,0,0), (0.0,0.0), (10.0,0.0))
        pygame.draw.line(canvas, (0,0,0), (10.0,0.0), (10.0,10.0))
        pygame.draw.line(canvas, (0,0,0), (0.0,10.0), (10.0,10.0))
        
        if self.render_mode == "human":
            # The following line copies our drawings from `canvas` to the visible window
            self.window.blit(canvas, canvas.get_rect())
            pygame.event.pump()
            pygame.display.update()

            # We need to ensure that human-rendering occurs at the predefined framerate.
            # The following line will automatically add a delay to keep the framerate stable.
            self.clock.tick(self.metadata["render_fps"])
        else:  # rgb_array
            return np.transpose(
                np.array(pygame.surfarray.pixels3d(canvas)), axes=(1, 0, 2)
            )
        
    def close(self):
        if self.window is not None:
            pygame.display.quit()
            pygame.quit()

测试代码 #

import gym

env = gym.make('MazeEnv-v0')

observation, info = env.reset()

for _ in range(1000):
    observation, reward, terminated, truncation, info = env.step(env.action_space.sample())
    
    if terminated:
        observation, info = env.reset()
        
env.close()

报错信息 #

Traceback (most recent call last):

  File ...\Code\Maze Game\test.py:13 in <module>
    observation, info = env.reset()

  File E:\anaconda3\lib\site-packages\gym\wrappers\time_limit.py:68 in reset
    return self.env.reset(**kwargs)

  File E:\anaconda3\lib\site-packages\gym\envs\classic_control\maze_game.py:48 in reset
    super().reset(seed=seed)

TypeError: reset() got an unexpected keyword argument 'seed'

问题原因及解决办法 #

问题原因 #

Gym v0.26.0版本中，基类gym.Env的reset()方法并不接受seed参数，你在自定义reset()方法里调用super().reset(seed=seed)时，给父类方法传入了它不支持的参数，因此触发TypeError。

解决步骤 #

1. 移除父类reset调用的seed参数：直接删掉super().reset(seed=seed)这行代码（Gym v0.26.0的Env.reset()本身没有实际逻辑，无需调用）。
2. 手动处理随机种子：如果你的环境需要设置随机种子，在自定义reset()里手动初始化种子即可，示例如下：

修改后的reset()方法：

def reset(self, seed=None, options=None):
    # 手动设置随机种子
    if seed is not None:
        random.seed(seed)
        np.random.seed(seed)
    
    # 初始化位置改为numpy数组，避免后续运算报错
    self._agent_location = np.array([0.0, 10.0], dtype=np.float32)
    self._target_location = np.array([10.0, 0.0], dtype=np.float32)
    
    observation = self._get_obs()
    info = self._get_info()
    
    if self.render_mode == "human":
        self._render_frame()
        
    return observation, info

另外补充：你当前代码中self._agent_location用的是普通列表，后续和direction（列表）相加没问题，但如果后续涉及numpy数组运算可能出错，建议统一改为numpy数组。

内容的提问来源于stack exchange，提问作者Daniel Rodriguez