Reinforcement Learning & Autonomous Driving Education with AWS DeepRacer at KNUT

Korea National University of Transportation (Chungju campus) successfully completed the “Reinforcement Learning and Autonomous Driving with AWS DeepRacer” program to nurture future autonomous driving talent. The program ran November 1–2 at SPACE K in the Business & Aviation Hall, with 22 students from diverse majors including computer science, automotive engineering, and electronics.

The two-day, 16-hour curriculum balanced theory and practice — from the fundamentals of AI, machine learning, and deep learning to how reinforcement learning works, tuning major algorithms such as PPO and SAC, and reward function programming. Participants trained models on the AWS cloud simulator and then deployed them to real vehicles to implement autonomous driving.

On the final day, a competition was held to test learning outcomes: the best simulation lap was 17.457 seconds in the individual event, and the best offline team lap was 8.113 seconds. The overall satisfaction rating was 4.85 out of 5, with participants noting that “reinforcement learning, which felt difficult, became easy and fun to learn through the familiar subject of autonomous driving.”

With 21 students completing the course and AI mobility ranked as the most-desired follow-up subject, the program is regarded as a meaningful stepping stone for strengthening regional students’ autonomous driving and AI capabilities.