Self Driving Car

Project description

Throughout the challenge, teams must develop, integrate, and test autonomous systems that can perceive the environment, make decisions, and control the vehicle accordingly. The ultimate objective is to complete the designated track autonomously and safely, demonstrating the practical application of advanced robotics, AI, and embedded systems in real-world traffic scenarios.

Context

The project intersects multiple technical domains including robotics, control systems, embedded programming, and machine learning. Success is measured by the vehicle’s ability to complete a closed-course track safely and autonomously, reflecting its readiness for real-world deployment.

Results

The most significant outcome of this project was the successful development and real-world deployment of a fully autonomous vehicle system capable of completing the RDW Self-Driving Challenge 2025 without human intervention — securing first place for the second year in a row with the first ever perfect run.
The system integrated YOLOv8-based models for traffic signs, traffic lights, and crosswalk detection, all deployed on an NVIDIA Jetson Orin. These were connected to a finite state machine that controlled behavior like stopping at STOP signs, yielding to pedestrians, and reacting to traffic lights.

All models were trained using both public datasets and custom footage from the track. The vehicle was validated through system testing, simulation, and live track runs, achieving >90% precision and recall. It successfully completed all tasks during the challenge with zero critical failures.

Our solution reached TRL level 6–7, having been tested in a relevant real-world setting with consistent, repeatable results. Key insights included the importance of dataset quality, fallback logic, and modular AI design.

About the project group

We are a group of ICT & Technology students, we all have done our specialization before in Smart Industries. This past semester we have worked on this project for the majority time, spending each day working on getting the vehicle ready for the competition and implementing autonomous modules.