Road Anomaly Detection and Localization for Connected Vehicle Applications

Abstract

<div class="section abstract"><div class="htmlview paragraph">Road anomalies pose significant challenges for on-road safety, ride comfort, and fuel economy. The recent advancement of Connected Vehicle technology has made it feasible to overcome this challenge by sharing the detected road hazards information with other vehicles and entities. However, localization accuracies of the detected road hazards are often very low due to noisy detection results and limited GPS sensor performances. In this paper, a cloud based data management system with in-vehicle and on-cloud data processing modules is presented for road hazards detection and localization. Stereo camera and a consumer-grade GPS sensor on a testing vehicle are used to detect road anomaly information, e.g., type, size, and location, where a novel in-vehicle data processing module is implemented based on Kalman Filter and Phase Adjustment. For hazards data shared from all connected vehicles, an on-cloud data processing module is designed to further improve anomaly localization accuracy based on clustering. The whole system was tested in a parking lot with potholes, debris, and road bumps. Experimental results show that the hazards localization accuracy could be significantly improved from 7.4m to 1.4m with 84% accuracy using the proposed system. The proposed real-time system could bring significant benefits for commercial vehicles, and transportation companies with improved safety and ride quality.</div></div>