Online State-to-State Time-Optimal Trajectory Planning for Quadrotors in Unknown Cluttered Environments

This paper introduces the first planner, called STAMINER (STAte-to-state tiMe-optImal memoryless planNER), which is able to real-time plan collision-free, local time-optimal trajectories in unknown and cluttered settings without the need for any maps or fused occupancy structures of the surrounding environment. Specifically, our method explores a library of state-to-state time-optimal trajectories in a memoryless collision-checking framework. It iteratively searches for the best trajectory with the longest projection on the direction to the goal. Results obtained from two different simulated cluttered scenarios demonstrate that our planner outperforms the state-of-the-art baselines concerning global finishing time. Furthermore, real-world flight trials were conducted to validate the effectiveness of our algorithm in an actual quadrotor. Finally, this paper also provides a mathematical proof of the solution characterization that is not available in the literature for the considered state-to-state time-optimal trajectory generation problem.