Superhuman Safe and Agile Racing through Multi-Agent Reinforcement Learning

2026-05-21

AI Summary

This work addresses the challenge of enabling autonomous systems to coordinate safely and efficiently in highly dynamic, real-world shared environments. It presents the first application of multi-agent reinforcement learning to physical quadrotor drone racing, integrating league-based self-play, aerodynamic downwash modeling, and an intrinsic coordination mechanism. Remarkably, the proposed approach achieves effective collision avoidance and coordination without explicit safety constraints. The method demonstrates zero-shot generalization to human–robot interaction and outperforms expert human pilots in multi-drone races at speeds exceeding 22 m/s, while reducing collision rates by 50% compared to a single-agent baseline.

Abstract

Autonomous systems have achieved superhuman performance in isolation or simulation, yet they remain brittle in shared, dynamic real-world spaces. This failure stems from the dominant single-agent paradigm for physical applications, where other actors are ignored or treated as environmental noise, preventing effective coordination. Here we show that multi-agent reinforcement learning provides the essential safety scaffolding required for real-world interaction. Using high-speed quadrotor racing as a high-stakes testbed, we train agents to navigate complex aerodynamic interactions and strategic maneuvering with a variable number of racers. Through league-based self-play, agents evolve sophisticated anticipatory behaviors, including proactive collision avoidance, overtaking, and handling multi-agent physical interactions, including aerodynamic downwash. Our agents outperform a champion-level human pilot in multi-player races at speeds exceeding 22 m/s, while simultaneously reducing collision rates by 50 % compared to state-of-the-art single-agent baselines. Crucially, training with diverse artificial agents enables zero-shot generalization to safer human interaction. These results suggest that the path to robust robotic co-existence lies not in isolated safety constraints, but in the rigorous demands of multi-agent interaction. Multimedia materials are available at: https://rpg.ifi.uzh.ch/marl

Problem

• multi-agent interaction
• autonomous systems
• real-world safety
• dynamic environments
• coordination

Innovation

• multi-agent reinforcement learning
• quadrotor racing
• aerodynamic interaction
• self-play
• zero-shot generalization