Optimal Trajectory Generation for Quadrotor Teach-and-Repeat

Abstract

In this letter, we propose a novel motion planning framework for quadrotor teach-and-repeat applications. Instead of controlling the drone to precisely follow the teaching path, our method converts an arbitrary jerky human-piloted trajectory to a topologically equivalent one, which is guaranteed to be safe, smooth, and kinodynamically feasible with an expected aggressiveness. Our proposed planning framework optimizes the trajectory in both spatial and temporal aspects. In the spatial layer, a flight corridor is found to represent the free space that is topologically equivalent with the teaching path. Then, a minimum-jerk piecewise trajectory is generated within the flight corridor. In the temporal layer, the trajectory is re-parameterized to obtain a minimum-time temporal trajectory under kinodynamic constraints. The spatial and temporal optimizations are both formulated as convex programs and are done iteratively. The proposed method is integrated into a complete quadrotor system and is validated to perform aggressive flights in challenging indoor and outdoor environments.