LiDAR priors slash 60-GHz beam search probing by 72%

Millimeter-wave 60-GHz links are highly directional and fragile — blockages kill connectivity, forcing devices to find non-line-of-sight (NLoS) paths via reflections off walls, floors, and ceilings. But scanning all possible beam directions to discover those paths is slow and energy-hungry, especially for IoT devices. A team led by Mohammed E. Eltayeb at the University of South Florida investigated whether LiDAR point clouds could short-circuit that process by providing a 'surface-aware prior' — ranking candidate propagation directions based on geometry and reflectivity, without assuming optical returns predict mmWave reflection strength.

The team's validation unfolded in three stages: first, they characterized how geometric and radiometric surface descriptors (like local 3-D planarity) vary with LiDAR acquisition geometry; then, they matched LiDAR data with 60-GHz measurements in an L-shaped corridor to see if descriptors correlate with real surface-mediated RF responses; finally, they ran a room-scale campaign using exhaustive TX-RX beam maps. The results show no deterministic mapping from LiDAR returns to RF power — but the two modalities do exhibit cross-modal association. In the room experiment, a local three-ring 3-D planarity descriptor ranked candidate directions well enough to retain a beam within 3 dB of exhaustive search at 74.5% of measured locations, while cutting RF beam-pair probing by 72% over the candidate search region.

This work matters because it decouples beam search from pure RF brute force: LiDAR-derived priors concentrate probing on the most promising directions, and RF measurements only finalize selection. That hybrid approach could significantly reduce training latency and power consumption in next-gen indoor 60-GHz IoT systems. While the paper stops short of claiming LiDAR can predict exact reflection paths, it establishes local surface structure as a practical prior for shrinking mmWave beam-search uncertainty.