Optimal sensing of urban road networks with large-scale ridesourcing vehicles

Abstract

The monitoring of the urban road network contributes to the efficient operation of the urban transportation system and the functionality of urban systems. Current practices of sensing urban road networks mainly depend on inductive loop sensors, roadside cameras, and crowdsourcing data from massive urban travelers (e.g., Google Map). These data have the drawbacks of high costs, limited coverage, or low reliability due to insufficient user penetration. This study investigates an innovative approach of drive-by sensing, which leverages large-scale ridesourcing vehicles (RVs) to monitor and infer the states of urban road networks. With RV traversing over road networks, we examine the RV fleet sensing performance based on the unique number of road segments explicitly visited, the sensing reliability as a result of repeated visits, and the information that can be implicitly inferred given explicit data and road network topology. We propose an optimal rerouting model to simultaneously maximize the sensing coverage and sensing reliability, which can be efficiently solved using a heuristic algorithm to guide the cruising trajectory of unoccupied RVs sequentially.To validate the effectiveness of the proposed model, comprehensive experiments and sensitivity analyses are performed using real-world RV data of more than 20,000 vehicles in New York City (NYC). Our approach is shown to cover up to 75.5% of all road segments which leads to an implicit coverage of 97.2%. More importantly, the coverage are obtained with at least 31.1% improvements in sensing reliability as compared to the baseline scenario.