Most of the autonomy work I have done commercially sits inside restricted repositories. ARCO exists to fill that gap with something I actually own and can share openly.
The scope covers the full planning and guidance pipeline: environment representation, graph and sampling-based path planning, and the controllers that follow a planned path. The progression mirrors the architecture of a real autonomy stack, built one layer at a time and validated at each stage before moving forward.
The implementation is Python, deliberately. My production work has been C++ and embedded systems. Here I want clean, well-documented scientific computing: readable algorithms, step-by-step visualizations that expose what each method actually does, and tests that make the assumptions explicit. The goal is a reference I can use in conversation with anyone trying to understand how these systems work.
Resources
- GitHub Repository. Full source code, demos, and documentation.
- Project Roadmap. Staged milestones from classical planning to ROS 2 integration.
Technical Overview
ARCO is organized around three layers that correspond to the standard decomposition of an autonomy stack.
- The mapping layer handles spatial representation.
- The planning layer searches or samples for feasible paths through that representation
- The guidance layer tracks the resulting path with feedback controllers (pure pursuit, PID, MPC).
Each layer is independently testable and documented, with visualizations designed for step-by-step inspection of algorithm behavior.