We are a University of Toronto student club that designs, builds and races small autonomous race cars. Members work across software and hardware, from perception and control to the car's electronics and chassis. New members are welcome at any experience level.
A 1/10-scale car that maps a track on its first lap and then races it on its own:
- Compute and sensing: an NVIDIA Jetson with an OAK-D stereo depth camera and an IMU
- Localization and mapping: ORB-SLAM3 visual SLAM, with our own CUDA acceleration of feature extraction and stereo matching
- Control: pure pursuit, Stanley, LQR and MPC track followers, plus learned policies trained in simulation, all behind a safety supervisor
- Embedded: Arduino firmware that drives the steering servo and ESC, with the RC transmitter always able to take over
| Repository | What it is |
|---|---|
| ORB_SLAM3 | Our ORB-SLAM3 fork with GPU ORB extraction, stereo matching and image pyramid |
| scale_racing_simulator | PyBullet and Gymnasium simulator for trying reinforcement learning policies |
| ResNet-Auto-Training | Track segmentation training with masks generated by SAM 2 |
| telemetry-platform-backend | Real-time telemetry ingestion and streaming for the car |
| website | Source for utasrteam.ca |
The car's onboard software and firmware are kept private to club members.
- Software: data engineering, machine learning, reinforcement learning, web development, software and optimization
- Hardware: mechanical, electrical, firmware
- Create a branch in the repository (
git checkout -b short-descriptive-name). - Commit small, focused changes with clear messages.
- Push the branch and open a pull request.
- Get a review from a lead before merging into
main.
- Website: utasrteam.ca
- Discord: join our server
- Instagram: @utasr_uoft
- Email: [email protected]
