In the field of robotics many different approaches ranging from classical planning over optimal control to reinforcement learning (RL) are developed and borrowed from other fields to achieve reliable control in diverse tasks. In order to get a clear understanding of their individual strengths and weaknesses and their applicability in real world robotic scenarios is it important to benchmark and compare their performances not only in a simulation but also on real hardware. The '2nd AI Olympics with RealAIGym' competition was held at the IROS 2024 conference to contribute to this cause and evaluate different controllers according to their ability to solve a dynamic control problem on an underactuated double pendulum system with chaotic dynamics. This paper describes the four different RL methods submitted by the participating teams, presents their performance in the swing-up task on a real double pendulum, measured against various criteria, and discusses their transferability from simulation to real hardware and their robustness to external disturbances.
View on arXiv@article{wiebe2025_2503.15290,
title={ Reinforcement Learning for Robust Athletic Intelligence: Lessons from the 2nd ÁI Olympics with RealAIGym' Competition },
author={ Felix Wiebe and Niccolò Turcato and Alberto Dalla Libera and Jean Seong Bjorn Choe and Bumkyu Choi and Tim Lukas Faust and Habib Maraqten and Erfan Aghadavoodi and Marco Cali and Alberto Sinigaglia and Giulio Giacomuzzo and Diego Romeres and Jong-kook Kim and Gian Antonio Susto and Shubham Vyas and Dennis Mronga and Boris Belousov and Jan Peters and Frank Kirchner and Shivesh Kumar },
journal={arXiv preprint arXiv:2503.15290},
year={ 2025 }
}