Mastering Diverse, Unknown, and Cluttered Tracks for Robust Vision-Based Drone Racing

2025-12-10

AI Summary

Visual navigation for high-speed drone racing in unknown, cluttered environments poses significant challenges—including tight coupling between gate passage and robust obstacle avoidance, susceptibility of policies to local optima, and ambiguous depth-map perception. Method: We propose a soft-hard collision two-stage reinforcement learning framework. It incorporates noise-augmented adaptive curriculum learning, an asymmetric Actor-Critic architecture, Lipschitz regularization for policy smoothness, and a track primitive generator to enhance generalization and motion stability. The method operates solely on depth-map input and supports joint simulation-to-real training. Results: Evaluated on computationally constrained quadcopters, our approach achieves highly agile flight, exhibits robustness to gate localization errors, and generalizes across diverse, partially unknown, cluttered environments—demonstrating successful deployment in both simulation and real-world settings.

Technology Category

• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Search and Optimization: Learning to Search
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments

Abstract

Most reinforcement learning(RL)-based methods for drone racing target fixed, obstacle-free tracks, leaving the generalization to unknown, cluttered environments largely unaddressed. This challenge stems from the need to balance racing speed and collision avoidance, limited feasible space causing policy exploration trapped in local optima during training, and perceptual ambiguity between gates and obstacles in depth maps-especially when gate positions are only coarsely specified. To overcome these issues, we propose a two-phase learning framework: an initial soft-collision training phase that preserves policy exploration for high-speed flight, followed by a hard-collision refinement phase that enforces robust obstacle avoidance. An adaptive, noise-augmented curriculum with an asymmetric actor-critic architecture gradually shifts the policy's reliance from privileged gate-state information to depth-based visual input. We further impose Lipschitz constraints and integrate a track-primitive generator to enhance motion stability and cross-environment generalization. We evaluate our framework through extensive simulation and ablation studies, and validate it in real-world experiments on a computationally constrained quadrotor. The system achieves agile flight while remaining robust to gate-position errors, developing a generalizable drone racing framework with the capability to operate in diverse, partially unknown and cluttered environments. https://yufengsjtu.github.io/MasterRacing.github.io/

Problem

• Generalizing drone racing to unknown, cluttered environments
• Balancing high-speed flight with robust collision avoidance
• Overcoming perceptual ambiguity in depth-based gate detection

Innovation

• Two-phase learning with soft and hard collision training
• Adaptive curriculum shifting from gate-state to depth input
• Lipschitz constraints and track-primitive for stability generalization