Urban Traffic Coulomb's Law: A New Approach for Taxi Route Recommendation

• Published in: IEEE transactions on intelligent transportation systems 2019-08, Vol.20 (8), p.3024-3037
• Main Authors: Lai, Yongxuan, Lv, Zheng, Li, Kuan-Ching, Liao, Minghong
• Format: Article
• Language: English
• Subjects: Algorithms; cruising route recommendation; Global Positioning System; Heuristic algorithms; Loads (forces); Optimization; Passengers; Public transportation; Recommender systems; Roads; Route selection; taxi trajectories; Taxicabs; Traffic force; Trajectories; Trajectory; Urban environments; urban traffic Coulomb’s law; Vehicles
• ISSN: 1524-9050; 1558-0016
• DOI: 10.1109/TITS.2018.2870990

Recently, an increased amount of effort has been focused on optimizing the selection of routes for taxis, as part of the development of smart urban environments, and the increase of the accumulated trajectory data sets. One challenging issue is to match and recommend appropriate cruising routes to taxis, as most taxis cruise on streets aimlessly looking for passengers. Drivers encounter lots of difficulty in optimizing their cruise routes and hence increasing their incomes, and such inability not only decreases their profit but also increases the traffic load in urban cities. In this paper, the concept of urban traffic Coulomb's law is coined to model the relationship between taxis and passengers in urban cities, based on which a route recommendation scheme is proposed. Taxis and passengers are viewed as positive and negative charges. It first collects useful information such as the density of passengers and taxis from trajectories, then calculates the traffic forces for cruising taxis, based on which taxis are routed to optimal road segments to pick up desired passengers. Different from existing route recommendation methods, the relationship among taxis and passengers are fully taken into account in the proposed algorithm, e.g., the attractiveness between taxis and passengers, and the competition among taxis. Moreover, real-time dynamics and geodesic distances in road networks are also considered to make more accurate and effective route recommendations. Extensive experiments are conducted on the road network using the trajectories generated by approximately 5,000 taxis to verify the effectiveness, and evaluations demonstrate that the proposed method outperforms existing methods and can increase the drivers' income more than 8%.