Evaluating Safety Risks in Pedestrian Facilities for Different Ride Modes: Application of a Deep Learning–Based 3D Scanning System

Personal mobility vehicles (PMVs) have seen a surge in popularity due to their convenience and ease of use. However, these vehicles are frequently utilized on sidewalks, bike lanes, and roads, thereby posing risks to both riders and pedestrians. The implementation of safety measures is vital to mitigate these risks. In many cases, even though, existing infrastructure and related inspection systems are not designed or maintained to accommodate PMVs, further elevating the risk of accidents. This study introduces an innovative system that employs a 3D scanner to collect point data and estimate path roughness. The system comprises three modules: Module 1—Shared path point cloud data (PCD) extraction; Module 2—Riding profile creation; and Module 3—Ride quality and risk assessment. A field test was conducted to validate the system’s performance at Purdue University. One key finding is the quantified probability of encountering severe drops or spikes in the testing area. Three different ride modes were tested, revealing varying levels of safety risk in the same area. In the no risk observed frequency data analysis, for example, skateboard (42%/577 segments), E-scooter (74%/953 segments), and wheelchair (96%/1,128 segments) risk percentages/frequencies were observed. The key implication derived from this analysis is that the quality assessment of a sidewalk surface should consider its physical attributes and the type of ride mode(s) used on it.