Dynamic ride-sourcing systems for city-scale networks - Part I: Matching design and model formulation and validation

The ubiquity of smart devices enables the foundation for emerging fast-growing ride-sourcing companies that challenge traditional taxi services. Two design aspects of on-demand mobility systems are: (i) the matching mechanism between idle ride-sourcing vehicles and passenger travel requests (i.e., vehicle–passenger matching) and (ii) the repositioning mechanism of idle vehicles. In this paper, we propose a macroscopic non-equilibrium dynamic model of ride-sourcing systems with capabilities of investigating the efficiency of vehicle–passenger matching and idle vehicle repositioning methods. A spatio-temporal vehicle–passenger matching method is introduced to determine dynamically and jointly the matching time instances and maximum matching distances to minimize passengers’ waiting time (i.e., from the travel request until the pickup) while considering the network congestion levels. Designing a controller for repositioning idle vehicles to balance vehicle supply and passenger travel demand based on the proposed model is scrutinized in the companion paper (Part II). The accuracy of the proposed model and the performance of the matching method under noticeable variations of traffic congestion and passenger travel requests are investigated with microsimulation. The results demonstrate the accuracy of the model in predicting the evolution of the number of ride-sourcing vehicles in different states (idle, transferred, dispatched, and occupied) and passengers (waiting and assigned) in each region of the network. Furthermore, the effectiveness of the proposed matching method is demonstrated by the decrease in the waiting times of ride-sourcing vehicles and passengers. •Long-distance vehicle–passenger matched pairs are discarded systematically.•The matching method is designed to occur at irregular intervals to achieve optimality.•A dynamic ride-sourcing model to track the states of vehicles and passengers.•The model is validated with high-fidelity microsimulation data.