Reliable Autonomous Landing and Hazard Avoidance of Drones Based on Elevation Map Landing Ability Analysis

The landing of drones in unknown environments poses a highly challenging task. This paper presents a safe and efficient LiDAR landing approach. The method leverages positioning data obtained from the perception information of LiDAR and the Inertial Measurement Unit (IMU), projecting point clouds into an elevation map coordinate system. Utilizing Bayesian generalized kernel elevation inference and an enhanced dynamic clearing algorithm, it predicts, fills, and dynamically updates sparse elevation maps to generate dense and comprehensive elevation maps. Subsequently, by analyzing terrain geometric parameters in the elevation map such as slope, roughness, and step height, a landing ability analysis map is generated. Then, employing a GPU-accelerated method, it swiftly identifies the safest landing position from the landing ability analysis map. Furthermore, addressing the potential degradation of LiDAR positioning near landing points, a strategy solely relying on single-frame point clouds for obstacle avoidance is proposed, achieving secure drone landings. This method has been validated across multiple complex simulation environments and real-world scenarios, showing excellent results.