Cloud-Edge LiDAR Segmentation (UoE Thesis)

Uncertainty-aware cloud-edge collaborative framework for real-time 3D LiDAR point cloud segmentation. Reduced cloud computation by 50%, maintained edge-to-cloud latency under 30ms, improved segmentation accuracy by 0.8pp over edge-only baseline. Provides a deployable real-time 3D perception paradigm for automotive-grade and low-altitude scenarios.