SMPLOlympics: Sports Environments for Physically Simulated Humanoids

TL;DR

SMPLOlympics introduces a unified, SMPL-compatible suite of physically simulated Olympic sports for humanoids, spanning single- and multi-person events and enabling motion-data transfer from human videos. It combines motion priors (e.g., PULSE, AMP) with simple, sport-specific rewards and a pipeline to extract demonstrations via TRAM/PHC, establishing a versatile benchmark for goal-conditioned RL and competitive self-play. The work provides baseline implementations, evaluation metrics, and curriculum insights, showing that strong motion priors plus straightforward rewards can yield human-like behaviors across diverse sports, with video data offering targeted gains. Together, SMPLOlympics lays groundwork for standardized evaluation and rapid prototyping of embodied agents in realistic athletic tasks, with clear paths for expansion and integration with frontier models.

Abstract

We present SMPLOlympics, a collection of physically simulated environments that allow humanoids to compete in a variety of Olympic sports. Sports simulation offers a rich and standardized testing ground for evaluating and improving the capabilities of learning algorithms due to the diversity and physically demanding nature of athletic activities. As humans have been competing in these sports for many years, there is also a plethora of existing knowledge on the preferred strategy to achieve better performance. To leverage these existing human demonstrations from videos and motion capture, we design our humanoid to be compatible with the widely-used SMPL and SMPL-X human models from the vision and graphics community. We provide a suite of individual sports environments, including golf, javelin throw, high jump, long jump, and hurdling, as well as competitive sports, including both 1v1 and 2v2 games such as table tennis, tennis, fencing, boxing, soccer, and basketball. Our analysis shows that combining strong motion priors with simple rewards can result in human-like behavior in various sports. By providing a unified sports benchmark and baseline implementation of state and reward designs, we hope that SMPLOlympics can help the control and animation communities achieve human-like and performant behaviors.

Figures (7)

• Figure 1: A collection of various sports environments for physically simulated humanoids.
• Figure 2: An overview of $\text{SMPLOlympics}$: we design a collection of simulated sports environments and leverage RL and human demonstrations (from videos or MoCap) as prior to tackle them.
• Figure 3: Qualitative results for high jump, javelin, golf, and hurdling. PPO and AMP try to solve the task using inhuman behavior, while PULSE can discover human-like behavior.
• Figure 4: Qualitative results for table tennis and tennis. PPO and AMP result in inhuman behavior; PULSE can use human-like movement but PULSE + AMP result in behavior specific to the sport.
• Figure 5: Learning curves on various tasks.