Design of autonomous obstacle avoidance system for multirotor unmanned aerial vehicles based on MID-360 LiDAR

Abstract

Autonomous navigation of multi-rotor Unmanned Aerial Vehicles (UAVs) in complex low-altitude environments remains challenging due to severe onboard computational constraints, sensor blind spots, and unpredictable dynamic obstacles. This paper presents a lightweight, integrated autonomous obstacle avoidance system for quadrotors utilizing a MID-360 solid-state LiDAR. To achieve robust state estimation, FAST-LIO2 is employed as a high-frequency odometry backend. For dynamic environments, an Unscented Kalman Filter(UKF) integrated with a Constant Turn Rate and Velocity (CTRV) motion model is proposed to detect, track, and predict the non-linear trajectories of moving obstacles in real time. For motion planning, an Euclidean Signed Distance Field (ESDF)-free gradient-based local planner is developed, which incorporates an adaptive safety inflation layer tailored specifically to compensate for the vertical field-of-view (FOV) blind zones of the MID-360 sensor during aggressive maneuvers. The entire system is deployed on an NVIDIA Jetson Orin Nano onboard compute platform. Extensive simulations and real-world flight tests demonstrate that the proposed system achieves ultra-low latency trajectory re-planning (within 10ms) and ensures robust collision-free flight against both static and dynamic hazards, exhibiting superior execution efficiency and safety.