rl_sar/README_CN.md

rl_sar

English document

强化学习的Gazebo仿真验证与UnitreeA1实物部署。"sar"代表"simulation and real"

准备

拉取代码（同步拉取子模块）

git clone --recursive https://github.com/fan-ziqi/rl_sar.git

如有更新：

git pull
git submodule update --remote --recursive

在任意位置下载并部署libtorch

cd /path/to/your/torchlib
wget https://download.pytorch.org/libtorch/cpu/libtorch-cxx11-abi-shared-with-deps-2.0.1%2Bcpu.zip
unzip libtorch-cxx11-abi-shared-with-deps-2.0.1+cpu.zip -d ./
echo 'export Torch_DIR=/path/to/your/torchlib' >> ~/.bashrc

安装 teleop-twist-keyboard

sudo apt install ros-noetic-teleop-twist-keyboard

编译

自定义代码中的以下两个函数，以适配不同的模型：

torch::Tensor forward() override;
torch::Tensor compute_observation() override;

然后到根目录编译

cd ..
catkin build

运行

运行前请将训练好的pt模型文件拷贝到rl_sar/src/rl_sar/models中

仿真

新建终端，启动gazebo仿真环境

source devel/setup.bash
roslaunch rl_sar start_env.launch

新建终端，启动控制程序

source devel/setup.bash
rosrun rl_sar rl_sim

新建终端，键盘控制程序

rosrun teleop_twist_keyboard teleop_twist_keyboard.py

实物

新建终端，启动控制程序

source devel/setup.bash
rosrun rl_sar rl_real

部分代码参考https://github.com/mertgungor/unitree_model_control