AI Olympics with RealAIGym

Date:

More information here

This talk addresses the success of Model-Based Reinforcement Learning in conquering the athletic intelligence challenge at IJCAI 2023, which revolved around the swing-up and stabilization of two underactuated 2-link robot systems. Our team secured first place in the competition by leveraging Model-Based Reinforcement Learning, competing against two teams that employed Model-Free Reinforcement Learning. This success underscores the efficacy of our approach in controlling underactuated robots.