RoboCar
A small car that drives itself, built by our campus team for Epitech's national RoboCar race, and 3rd at the final in Paris. It sees the track through a single camera and decides on its own, with an AI model for tricky turns or a lighter algorithm when every bit of computing power counts.
Team
6 people
Context
Epitech RoboCar programme, one team per campus
Status
In progress
Context
RoboCar started as a challenge from UTAC to Epitech students. Each campus enters one team, which builds a self-driving car and races it against the other campuses.
What was built
- Two driving modes from the same camera: an image-recognition model trained with supervised learning, more resilient in complex turns, and a ray-casting algorithm on the camera's output that uses far fewer resources
- Decision algorithms written in C that turn the camera analysis into throttle and steering
- Our own ESC library in C (wesc): a rewrite of the VESC UART protocol for better performance than the standard tooling, soon to be open-sourced
- Hardware integration: NVIDIA Jetson Nano, OAK-D Lite camera, VESC electronic speed controller and steering servo
Results
3rdplace at the RoboCar final in Paris