File size: 1,276 Bytes
15b8951 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 | services:
# Driver + navigation (Nav2/SLAM). Default profile — `docker compose up go2`.
go2:
build:
context: ..
dockerfile: docker/Dockerfile
image: go2-nav-vla:latest
environment:
- ROBOT_IP=${ROBOT_IP:?set ROBOT_IP to your Go2's LAN address}
- ROBOT_TOKEN=${ROBOT_TOKEN:-}
- CONN_TYPE=${CONN_TYPE:-webrtc}
- DISPLAY=${DISPLAY:-}
volumes:
- /tmp/.X11-unix:/tmp/.X11-unix:ro # X11 forwarding, e.g. for rviz2
- ${HOME}/.Xauthority:/root/.Xauthority:rw
network_mode: host # required: ROS 2 DDS discovery + Go2 WebRTC live on the LAN
privileged: true # required: joystick/input device access
devices:
- /dev/input:/dev/input
restart: unless-stopped
command: ["ros2", "launch", "go2_navigation", "bringup.launch.py"]
# Local VLA brain (Qwen2.5-VL) + web console, port 8001. Opt-in:
# docker compose --profile vla up vlm
# Requires the NVIDIA Container Toolkit on the host for GPU access.
vlm:
extends: go2
profiles: ["vla"]
working_dir: /ros2_ws/src
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]
command: ["./vlm/scripts/run_demo1.sh"]
|