Incremental recommendation system for large-scale taxi fleet in smart cities

Modern urbanization is demanding high-efficiency taxi services to satisfy the request of taxi users as well as the expectation of drivers. Customers desire to wait the minimum time before finding a taxi, while drivers aim to maximize their profits by speeding up their customer hunting. In this paper, we propose to exploit the benefits of vehicular social network by enabling data exchange and more cooperation among regular taxi drivers. The objective is to design a smart recommendation sys-tem allowing drivers to optimize three key performance metrics: number of pickups, customer waiting time, and vacant travel distance. The proposed recommendation starts by efficiently estimating the future customer demands in the different region of interest. Then, it proposes an optimal taxi-to-region matching according to the current location of each taxi and the requested demand. Finally, an optimized routing algorithm is proposed to allow taxi drivers reach their destinations faster. Our simulation model is applied on the borough of Manhattan, NYC, and is validated with realistic data. The selected results show that significant performance gains can be achieved as compared to the traditional scenario.