Learn How Self-Driving Cars Think, in Python

What it does This is a curated set of self-contained Python simulations covering the full autonomy pipeline: localization, mapping, path planning, path tracking, and perception. Each script implements a single algorithm—like an Extended Kalman Filter, Hybrid A*, or Model Predictive Control—and renders it as a Matplotlib animation so you can see the math work. The author intends it as a textbook-in-progress for beginners who want to learn through isolated, runnable examples.

The interesting bit The author is building a future technical book around these exact snippets, which explains the deliberate, pedagogical isolation of each module. Rather than a monolithic simulator, you get a bag of decomposed puzzles: run extended_kalman_filter_localization.py and you get just that, with no spooky hidden state. It is autonomy education by dissection.

Key highlights

• Covers the full stack: EKF/UKF/Particle Filter localization, grid/NDT mapping, A*/RRT*/D* planning, Pure Pursuit/LQR/MPC tracking, and basic perception.
• Dependencies are minimal: only Python, NumPy, SciPy, and Matplotlib.
• CI tested across Linux, Windows, and macOS.
• Includes ongoing design documents, though the README admits they are still incomplete.

Caveats

• Design documents are explicitly incomplete and being updated.
• Simulations are 2D Matplotlib animations; there is no mention of real-world vehicle interfaces or sensor hardware.
• The project is strictly educational study material, not a production framework.

Verdict Grab this if you are studying robotics, prepping for autonomy interviews, or teaching the basics of self-driving algorithms. Look elsewhere if you need production C++ pipelines, ROS integration, or real sensor drivers.