Real-Time Optimization-Based Quadrotor Trajectory Generation with Kinodynamic Constraints in Unknown Environments

Abstract

Indoor disaster relief and rescue missions require quadrotors to fully exploit their maneuverability in real-time. However, the computational complexity induced by the underactuated kinodynamics conflicts with the rapid replanning requirement. For agile trajectory planning in cluttered and unknown environments, we propose a real-time optimization-based quadrotor trajectory generation method that integrates kinodynamic constraints in both trajectory search and trajectory optimization phases to fully exploit maneuverability. To further improve efficiency, we introduce a waypoints selection strategy to reduce the computational burden of kinodynamic trajectory optimization by transforming obstacle avoidance constraints into waypoint constraints, thereby enabling safe trajectory optimization in real-time. Specifically, kinodynamic trajectories are searched under kinodynamic constraints, providing reliable initial values for subsequent numerical optimization. Nextly, a waypoints selection algorithm, based on an estimation of trajectory variation during optimization, is introduced to preserve the obstacle-avoidance properties obtained during the search phase by limiting the variation with waypoint constraints. Finally, trajectory is segmented by waypoints with fixed time intervals each segment and then optimized under kinodynamic constraints, ensuring real-time optimization at the cost of time allocation optimality. We evaluated our method through simulation and experimentally validate its performance in cluttered and unknown environments. The competence of proposed method is also validated in real-world experiments.