恒源云服务器ubuntu系统Python3.8安装mujoco200、mujoco-py2.0.2.8

🔧 一、安装前准备

1. 安装系统依赖

   sudo apt update
   sudo apt install build-essential libgl1-mesa-glx libglew-dev patchelf libosmesa6-dev libglfw3-dev python3-dev
   

   确保GCC编译器、OpenGL驱动等基础工具完整安装，避免后续编译失败。

2. 获取MuJoCo许可证

   • 访问官网 https://www.roboti.us/license.html。
   • 下载许可证文件 mjkey.txt，需确认安装目标设备。

📂 二、安装MuJoCo核心文件

1. 下载与解压

   wget https://www.roboti.us/download/mujoco200_linux.zip -P ~/
   unzip ~/mujoco200_linux.zip -d ~/.mujoco
   mv ~/.mujoco/mujoco200_linux ~/.mujoco/mujoco200
   cp mjkey.txt ~/.mujoco/  # 复制许可证到主目录
   cp mjkey.txt ~/.mujoco/mujoco200/bin/  # 复制到bin目录
   cp mjkey.txt ~/.mujoco/mujoco200/  # 复制到目录
   

   解压后必须重命名文件夹为 mujoco200，否则路径识别可能失败。

2. 验证核心安装

   cd ~/.mujoco/mujoco200/bin
   ./simulate ../model/humanoid.xml
   

   若显示橘黄色人形模型动画，说明MuJoCo核心安装成功。

⚙️ 三、配置环境变量

编辑 ~/.bashrc 或 ~/.zshrc，末尾添加：

echo 'export MUJOCO_PY_MUJOCO_PATH=$HOME/.mujoco/mujoco200' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.mujoco/mujoco200/bin' >> ~/.bashrc
echo 'export MUJOCO_GL=egl' >> ~/.bashrc
source ~/.bashrc

生效配置：

source ~/.bashrc

🐍 四、安装mujoco_py（Python绑定）

1. 创建Conda环境（推荐）

   conda create -n mujoco_env python=3.8  # Python 3.8兼容性最佳
   conda activate mujoco_env
   

2. 安装Cython（一定要这个版本，否则会报错）

pip install cython==0.29.24

1. 安装mujoco_py
   一定要这个版本

   pip install mujoco-py==2.0.2.8  # 指定兼容的版本
   

   若安装失败，检查是否遗漏 libosmesa6-dev 或 patchelf 等依赖。

✅ 五、验证完整安装

import os
os.environ['LD_LIBRARY_PATH'] += ':/root/.mujoco/mujoco200/bin'
import mujoco_py

import mujoco_py
import os

# 1. 取 tuple 第 0 个元素
# mj_path = mujoco_py.utils.discover_mujoco()[0]

# 2. 或者直接解压（如果返回的一元组）
mj_path,_ = mujoco_py.utils.discover_mujoco()

xml_path = os.path.join(mj_path, 'model', 'humanoid.xml')
model = mujoco_py.load_model_from_path(xml_path)
sim = mujoco_py.MjSim(model)

若无报错且数据更新，表明mujoco_py集成成功。

✅程序报错1：

Import error. Trying to rebuild mujoco_py.
---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
Cell In[1], line 3
      1 import os
      2 os.environ['LD_LIBRARY_PATH'] += ':/root/.mujoco/mujoco200/bin'
----> 3 import mujoco_py

File /usr/local/miniconda3/lib/python3.8/site-packages/mujoco_py/__init__.py:3
      1 #!/usr/bin/env python
      2 import os
----> 3 from mujoco_py.builder import cymj, ignore_mujoco_warnings, functions, MujocoException
      4 from mujoco_py.generated import const
      5 from mujoco_py.mjrenderpool import MjRenderPool

File /usr/local/miniconda3/lib/python3.8/site-packages/mujoco_py/builder.py:506
    502     functions.mj_activate(key_path)
    505 mujoco_path, key_path = discover_mujoco()
--> 506 cymj = load_cython_ext(mujoco_path)
    509 # Trick to expose all mj* functions from mujoco in mujoco_py.*
    510 class dict2(object):

File /usr/local/miniconda3/lib/python3.8/site-packages/mujoco_py/builder.py:106, in load_cython_ext(mujoco_path)
    104     if mod is None:
    105         cext_so_path = builder.build()
--> 106         mod = load_dynamic_ext('cymj', cext_so_path)
    108 return mod

File /usr/local/miniconda3/lib/python3.8/site-packages/mujoco_py/builder.py:125, in load_dynamic_ext(name, path)
    123 """ Load compiled shared object and return as python module. """
    124 loader = ExtensionFileLoader(name, path)
--> 125 return loader.load_module()

ImportError: libglewosmesa.so: cannot open shared object file: No such file or directory

解决方案：如果 libglewosmesa.so 只在 mujoco 目录，直接链接到系统库路径：

sudo ln -sf /root/.mujoco/mujoco200/bin/libglewosmesa.so /usr/lib/libglewosmesa.so
sudo ln -sf /root/.mujoco/mujoco200/bin/libglewosmesa.so /usr/lib/x86_64-linux-gnu/libglewosmesa.so

（部分系统 /usr/lib64 或 /usr/local/lib 更合适，可按需调整。）

✅程序报错2：

next_obs, reward, done, info = self.env.step(action)
obs, reward, done, info = self.env.step(action)
ValueError: not enough values to unpack (expected 5, got 4)

建议安装兼容的Gym版本：

pip install gym==0.23.1