RESC: A Reinforcement Learning Based Search-to-Control Planning Framework for Agile Quadrotors

Agile flight in complex environments remains challenging for motion planning methods, which often rely on precomputed trajectories and simplified dynamics, limiting performance during aggressive maneuvers. While trajectory optimization methods typically decouple planning and control, optimizing trajectories separately from control dynamics, further constraining their ability to generate aggressive and feasible motions. To address these challenges, we propose an enhanced Search-to-Control planning framework that combines visibility-based path searching with RL-driven low-level control generation, eliminating explicit trajectory representation and bridging planning and enabling direct, dynamics-aware control. Our method first extracts control points from collision-free paths using a proposed heuristic search, which are then refined by an RL policy to generate low-level control commands for the quadrotor controller, utilizing reduced-dimensional obstacle observations for efficient inference with lightweight neural networks. We validate the framework through simulations and real-world experiments, demonstrating improved time efficiency and dynamic maneuverability compared to existing methods, while confirming its robustness and applicability.