Abstract

---

We present CrashSniffer, an anchor-free pedestrian collision prediction system for personal mobility (PM) vehicles such as e-scooters and e-bikes using Ultra-Wideband (UWB) sensing. CrashSniffer introduces a Virtual Antenna Array technique that harnesses the natural motion of PM vehicles to enhance localization accuracy without relying on external infrastructure. Coupled with a least-squares estimator, CrashSniffer enables pedestrian tracking under the challenging vehicle occlusion scenario. We then propose a Mobility-Aware Collision Prediction algorithm that considers pedestrian trajectories and directional intent to predict collisions. Field experiments demonstrate that CrashSniffer outperforms GPS and baseline UWB-based methods in pedestrian localization and collision prediction in realistic scenarios. Our scalable system offers a practical pathway to safer PM operation in pedestrian environments.

Publications

---

CrashSniffer: UWB-Based Anchor-Free Pedestrian Collision Prediction for Personal Mobility Vehicles
PDF

People

---