Docker Tutorial for DIAL-MPC

This repository contains the Dockerfile of the DIAL-MPC. DIAL-MPC is an impressive framework for legged robot full-order torque-level control with both precision and agility in a training-free manner. You can find the DIAL-MPC in here.

Requirenments

Caution

You need a Linux or Ubuntu system with Nvidia-driver (>=525.60.13) to support CUDA 12.*

Docker and Nvidia-toolkit installation

Step 1. Docker install

curl -fsSL https://get.docker.com -o get-docker.sh
sudo bash get-docker.sh

Step 2. Docker Postinstall(Follow here)

Step 3. Docker Nvidia Toolkit(Follow here)

Step 4. (Optional) Install xhost

sudo apt install x11-xserver-utils

Build Docker Env

git clone https://github.com/XuXinhangNTU/dial-mpc-docker
cd dial-mpc-docker
docker build -t dial-mpc .

First Time Run (Create the container)

xhost +
docker run -it -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=unix$DISPLAY --gpus all --name <Your Name> dial-mpc

Second Time Run

xhost +
docker start <Your Name> && docker exec -it <Your Name> bash

Tips

The vscode plugin "Dev Containers" is very easy to use when you do docker development.

Notice

Caution

If the demo looks odd, check your GPU usage! RTX 3090 is tested.