From Simulated to Real Test Environments for Self Driving Cars

In my master thesis in Artificial Intelligence, I studied testing in the field of self-driving cars through a small-scale car and simulator.

Through the use of CycleGAN, I propose a method to estimate the Cross-Track Error in the real world (important testing metric already in use for simulators) and use it to assess whether offline and online testing for self-driving cars yields similar results, both in a real and simulated environment.

Given the enthusiasm that me and my co-supervisor had towards this small-scale car, we even organized the first FormulaUSI event! The goal of the event was to educate participants on Artificial Intelligence while racing self-driving small-scale cars. We had much fun organizing the event, and I have personally grown by such an experience.

My master thesis can be downloaded at this link.
Here's the links to the FormulaUSI competition website and highlights.