Outracing Champion Gran Turismo Drivers With Deep Reinforcement Learning
Toronto Machine Learning Series (TMLS) via YouTube
Overview
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Discover how deep reinforcement learning was used to create an AI agent capable of competing against world-class e-sports drivers in Gran Turismo. In this 37-minute talk from the Toronto Machine Learning Series, Sony AI Senior Research Scientist Varun Raj Kompella explains the development of Gran Turismo Sophy. Learn about the challenges of training an AI to make real-time decisions in a complex physical system while interacting with humans and adhering to sportsmanship rules. Explore the combination of model-free deep reinforcement learning algorithms with mixed-scenario training to create an integrated control policy balancing speed and tactics. Gain insights into constructing reward functions that enable competitive AI agents to respect human norms in imprecisely defined domains. Witness how these techniques can be applied to control complex dynamical systems beyond racing simulations.
Syllabus
Outracing Champion Gran Turismo Drivers With Deep Reinforcement Learning
Taught by
Toronto Machine Learning Series (TMLS)