PythonRobotics: Python Code Collection and Textbook for Robotics Algorithms
NewPythonRobotics (Atsushi Sakai, since 2016; 30.8k+ stars) is a Python code collection and an online textbook of robotics algorithms. Its design goals are threefold: each algorithm is easy to read so the basic idea is clear; only widely used, practical algorithms are selected; and dependencies are kept minimal (NumPy, SciPy, Matplotlib, cvxpy).
The codebase is organized into algorithm families, each shipped as a runnable script with an animation GIF: Localization (EKF, particle filter, histogram filter, GPS/IMU fusion), Mapping (grid maps, ray casting, k-means clustering), SLAM (ICP, FastSLAM), Path Planning (Dijkstra, A*, D* Lite, RRT / RRT*, state lattice, PRM, LQR-RRT*, quintic polynomials, Reeds-Shepp, Frenet-frame optimal trajectory), Path Tracking (Stanley, rear-wheel feedback, LQR, MPC), Arm Navigation, Aerial Navigation, and Bipedal planning.
It is paired with a full Sphinx documentation (the "textbook") covering the mathematical background of each method, and it has an associated arXiv paper (Sakai et al., 2018; arXiv:1808.10703) describing the project's design and educational philosophy. For physics-AI practitioners, PythonRobotics is the canonical non-learning baseline to benchmark classical pipelines against learned planners and controllers before reaching for data-driven methods.