Autonomous Obstacle Avoidance and Navigation Technologies for Unmanned Aerial Vehicles Based on LiDAR: A Review

Abstract

Light Detection and Ranging (LiDAR) has become a benchmark sensing modality for autonomous unmanned aerial vehicle (UAV) navigation in GPS-denied and obstacle-dense environments such as forests, urban canyons, and indoor structures. This paper presents a structured review of the LiDAR-based UAV autonomy pipeline, spanning raw point-cloud processing, three-dimensional environment representation and mapping, simultaneous localization and mapping (SLAM), multi-sensor fusion, and real-time obstacle avoidance and trajectory planning. Reactive geometric methods, volumetric and distance-field mapping frameworks, tightly coupled LiDAR–inertial and LiDAR–inertial–visual odometry systems, gradient- and sampling-based trajectory optimizers, and learning-based end-to-end policies are compared with respect to computational cost, robustness, and applicability to resource-constrained micro-UAV platforms. The review further synthesizes current technical bottlenecks, including onboard computational limits, LiDAR performance degradation under adverse atmospheric conditions, and the difficulty of tracking fast-moving dynamic obstacles, as well as emerging research directions such as solid-state LiDAR integration, kinodynamic trajectory optimization, multi-sensor fusion (including radar- and event-camera-assisted schemes), learning-based exploration and foundation-model-based control, multi-UAV collaborative mapping, and simulation-to-reality transfer. The synthesis indicates that LiDAR remains a strong perceptual backbone for UAV autonomy, but that state-of-the-art systems increasingly combine it with inertial, visual, radar, and learning-based components rather than relying on LiDAR in isolation.