CyberPhysics

CyberPhysics is a container-first workspace for robotics, simulation, and ML applications. It keeps reusable application images in applications/, robot-specific calibration and configuration in systems/, and runnable multi-container stacks in compositions/.

The repository is designed for Ubuntu 24.04 and ROS 2 Jazzy unless an application README says otherwise.

Repository Layout

• applications/: reusable Dockerized tools such as ROS 2, Gazebo, Isaac Sim integrations, nvblox, Depth Anything, NiceGUI, DemNav, WildNav, and robot-specific bridges.
• systems/: hardware or simulation setups, including URDF/Xacro files, calibration, controller YAML, topic conventions, and robot-specific navigation config.
• compositions/: Docker Compose files that combine applications into working systems.

Keep large generated assets out of git. Docker images may download model weights, simulator assets, or example datasets during build; runtime maps should be downloaded or generated outside the repository.

Install

./install.sh

Build Images

Build from the applications/ directory with make build_<app_name>:

make -C applications build_ros2
make -C applications build_racecarneo
make -C applications build_depthanything

For ARM64 targets such as Raspberry Pi or Jetson:

cd applications
make target_arm64 build_ros2

Run A Composition

docker compose -f compositions/marimo.yaml up

The RACECAR Neo simulation and navigation stack is a representative full pipeline:

docker compose -f compositions/racecarneo.yaml up

That stack runs the headless MIT RACECAR Neo simulator, publishes ROS 2 camera/lidar/IMU/odometry topics, estimates metric monocular depth with Depth Anything, builds nvblox TSDF/ESDF slices, runs Nav2, and serves a NiceGUI waypoint interface at http://localhost:8080.

Observe ROS Data

Foxglove can be launched through a compose stack or on the host:

docker compose -f compositions/foxglove.yaml up
foxglove-studio

For nvblox visualization in Foxglove, install NVIDIA's Foxglove extension on the host:

git clone https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_nvblox.git
cd isaac_ros_nvblox/nvblox_foxglove
npm install
npm run local-install

The extension expects Node/npm compatible with NVIDIA's instructions; Node 20.10.0 has worked for this setup.

Development Conventions

• Put reusable software in applications/<name>.
• Put robot-specific configuration in systems/<system>.
• Put runnable stacks in compositions/<stack>.yaml.
• Prefer headless simulation containers that publish ROS 2 topics.
• Do not commit downloaded models, simulator assets, maps, logs, or bags.

Real-Time Notes

Pin latency-sensitive containers to dedicated CPU cores when needed:

docker run --cpuset-cpus="0,1" -it your_image your_command

services:
  your_service:
    image: your_image
    cpuset: "0,1"

Reserve the cores on the host before launching containers:

cset shield --cpu 0,1 --kthread on

Restore the host when finished:

cset shield --reset