How modular-autonomous-robotic-systems/neurorobotics-ws works
Project type and purpose
• This project is not a classic web or mobile app but a Docker-based robotics development environment. It automates building and running multiple Docker images/containers that host ROS-based SLAM, calibration, visualization (Pangolin), and related toolchains. The “djinn” script is the entry point for installing prerequisites, building images, and starting containers for various subsystems (NRT, Kalibr, Skynet-ROS, etc.).
File structure
• Top‐level scripts/ directory (djinn, install.sh, start.sh, build.sh, etc.) implements the orchestration CLI.
• docker/ contains subdirectories for each Docker image:
• base/ – generic base environment
• ros/ – ROS over base
• nrt/ (camera, pangolin) – real-time modules
• ext/ contains calibration toolchains (kalibr, allan_variance_ros) each with its own Dockerfile.
• config/ holds JSON settings for different hardware/environments and an RViz config.
• ros_ws/ is presumably where ROS packages live or get checked out by scripts.
• .gitmodules indicates some repos are embedded as submodules (likely ros packages).
README for additional insights
• “djinn install --docker --ssh --git” sets up host dependencies, clones repos, builds all Docker images.
• “djinn init docker [nrt|kalibr]” builds selected images & starts containers.
• “djinn up nrt”/“djinn exec nrt bash” show the typical workflow of interacting with containers.
• External dependencies (BFX, repos in ext/) must be manually placed.
Detail without overthinking
• Explicitly show:
• Build phase (scripts → Dockerfiles → images)
• Runtime phase (containers ← volume mounts ← host, container interconnections)
• External dependencies (Git repos, X server)