Skip to main content
Back

Donkeycar 5.3.0 Lets Students Build Self-Driving RC Cars on a Budget

A Python framework turns a hobby RC car into a self driving robot, with recent updates adding Raspberry Pi 5 support and simplified TFLite deployment.

Repo, Full Access
  • Donkeycar is an MIT-licensed Python library for building small-scale self driving cars on Raspberry Pi.
  • Version 5.3.0 adds Raspberry Pi 5 compatibility and simplified TFLite inference support.
  • Supports three autopilots: deep learning (TF, TFLite, PyTorch), GPS path-follow, and classical computer vision.
  • Modular parts pipeline handles cameras, lidar, controllers, motors, and data logging in one vehicle loop.
  • Ships with a simulator so you can train and race without owning hardware.
  • Powers the DIY Robocars race series with online and in-person events.

Self-driving stacks tend to live inside billion-dollar autonomy programs. Donkeycar takes the other route: a Python library that turns a hobby-grade RC car into a small-scale autonomous vehicle you can build on a kitchen table, train in the cloud, and race against other hobbyists. The project just crossed 3.5k stars and cut a new 5.3.0 release that modernizes its hardware and inference stack.

The library is minimalist and modular, aimed at hobbyists and students who want fast experimentation and low-friction contributions. High schools and universities use it for teaching and research, and it ships with a graphical interface and a simulator so you can experiment with self-driving before soldering anything.

Donkeycar prototype with Raspberry Pi, camera, and RC car chassis

What landed in 5.3.0

The latest release is a cleanup pass rather than a rewrite, but the changes matter for anyone assembling a car today. Highlights from the changelog:

  • Disabled camera alignment for Pi5 compatibility, unblocking builds on the newest Raspberry Pi.
  • Simplified TFLite support, plus a follow-up bug fix in the TFLite PR, making it easier to run quantized models on the Pi.
  • Reorganized and cleaned up the config file, which is the main file users edit to change their car.