Bridging Performance and Generalization in Reinforcement Learning for Agile Flight

2026-06-25

AI Summary

Existing reinforcement learning approaches struggle to simultaneously achieve high-speed performance and strong zero-shot generalization in agile drone racing. This work proposes a training framework that integrates task-aware policy switching with a physics-informed, procedural racetrack generator, significantly enhancing generalization without compromising racing speed. For the first time in end-to-end vision-based control, the method demonstrates robust zero-shot generalization without requiring test-time adaptation, achieving a 7.4× improvement in generalization performance over the current state-of-the-art across diverse real-world tracks never seen during training. The approach maintains competitive flight speeds and has been validated through both simulation and real-world experiments.

Abstract

Autonomous drone racing is a fundamentally challenging regime for autonomous aerial robots, requiring time-optimal control while operating under persistent actuation saturation. While reinforcement learning (RL) has achieved human-level performance in this domain, current methods fail to generalize; policies trained on specific environments often crash immediately in unseen configurations. This failure reflects the intrinsic difficulty of zero-shot generalization in agile flight, arising from high-dimensional task variation and the tight coupling between safety and performance at high speeds. Existing approaches that improve generalization impose a substantial cost on flight speed: control policies must significantly degrade performance to achieve even modest levels of generalization. In this work, we propose a framework for zero-shot generalization in agile flight for RL-based drone racing. By combining task-aware switching based on learning progress with a physically informed procedural track generator, the framework produces a fast and robust generalist policy without test-time adaptation. Our method achieves strong zero-shot performance across a wide range of unseen racetracks in the real world, demonstrating a 7.4x improvement in generalization over the state-of-the-art approaches, while maintaining competitive racing speeds. We validate our method's results in both simulation and real-world settings, including a challenging vision-based, end-to-end control setting that operates without explicit state estimation, where all prior approaches fail to generalize.

Problem

• zero-shot generalization
• agile flight
• reinforcement learning
• drone racing
• performance-generalization trade-off

Innovation

• zero-shot generalization
• reinforcement learning
• agile flight
• procedural track generation
• task-aware switching