Choosing a roadside LiDAR for traffic monitoring usually means comparing spec sheets rather than controlled evidence. The MR-LiDAR benchmark, described in a new arXiv preprint, tests four configurations—16, 32, 80, and 128 beams—against the same dynamic vehicle and vulnerable road user targets under matched daytime, nighttime, and occlusion conditions. Separate subsets vary mounting height, target speed, and curved-road geometry.

Among the main findings, 16- and 32-beam sensors lose target points earlier as distance grows, and pedestrians and cyclists are captured more sparsely than motor vehicles. Nominal beam count alone does not explain roadside scanning capability: the 80- and 128-beam configurations performed comparably, which the authors link to their nearly identical downward beam layouts. A counterfactual uniform 80-beam layout would yield only about one third of the effective beam count of the actual non-uniform configuration.

Illumination and the tested speed range had little influence on target-point-count patterns, whereas mounting height and near-field road geometry could substantially alter configuration ordering. The paper introduces effective beam count and a point-based Occlusion Ratio, and it provides an empirical minimal-sufficient configuration matrix for bounded deployment guidance.