Estimating Vehicle Speed with Consumer Grade Mobile LiDAR

Abstract

LiDAR (Light Detection and Ranging) is an emerging sensor technology that measures the time of flight of an emitted laser to measure the depth of surrounding objects. While historically LiDAR has been relegated to industrial and research spaces due to its prohibitive pricing and large form factor, recent developments have made it possible to include short range LiDAR on mobile devices. It is reasonable to postulate that technological developments will enable further adoption and performance enhancements. The high accuracy and resilience of LiDAR proves critical in providing autonomous vehicles robust information on their surroundings. But what if this capability could also be used to enhance the safety of the estimated 50 million commuters using bicycles, e-bikes, and scooters - micromobility riders - sharing the road, often dangerously, with cars? We explore the feasibility of reliably and accurately determining vehicle speed using a LiDAR-enabled mobile device mounted to a bicycle. We implemented an iOS application to gather real-world driving data, created a vehicle track matching algorithm to ascertain ground truth speed, and evaluated both a heuristic and a learned approach to estimate speed on LiDAR data.