Client: Dr. Animnesh Yadav, Ohio University EECS.
Team members (Pictured left to right):
- Team leader: Tin Vuong
- Documentation Manager: Elijah Beard
- Quality Assurance: Viet Huy Vu
- Release Manager: Minh Le
This project trains a reinforcement learning agent to drive a toy car by following a black line. We develop and validate policies in simulation, then deploy them to a Raspberry Pi-powered car for real-world testing. The project was awarded 1st place at the 2026 Ohio University Student Research Exposition.
Tech stack:
- Python
- Stable-Baselines3 and SB3-Contrib
- Gymnasium
- Pygame
- Raspberry Pi
- Toy car hardware (sensors + motor control)
We have explored camera-based approaches as alternatives to sensor-based line following. These experiments are tracked in separate branches:
-
huy_sim: A vision-based variant of the line follower that replaces the sensor array with camera input. The agent learns to identify and follow the black line from a camera.
-
old-donkeycar: An early attempt to train a model for general road driving (not limited to a single line). The model overfitted to its training environment and could not generalize to new tracks.
We use the MIT license.
You can find the installation instructions here.
You can find the user manual here.