Human-Guided Safe and Efficient Trajectory Replanning for Unmanned Aerial Vehicles

Abstract

Safe and efficient local trajectory replanning is essential for the navigation of unmanned aerial vehicles (UAVs). Take the quadrotor as an example, most research works focus on the static or fully mapped environment. Flying in a dynamic environment for autonomous quadrotors is still a tricky problem. However, with the emergence of first-person-view Drone Racing in recent years, professional human pilots have shown highly-skilled techniques for navigating quadrotors to avoid collisions at high speed. Therefore, this work uses the intelligence of human users in perception and decision-making and proposes a human-guided trajectory replanning (HTP) system for the safe and efficient flight operation of quadrotors. A non-constraint optimization problem is formulated, and human guidance is designed as one term of the cost functions. The proposed approach is validated in the AirSim simulation environment. The result shows that HTP saves optimization time by 58% compared with the non-human guidance (Non-HG) baseline. In addition, the HTP can assist quadrotors to pass the specified target at a higher speed and comply better with human preferences than the Non-HG approach.