Equilibrium analysis, pricing, and route choice for shared on-demand mobility services

Abstract

The advances in global positioning systems (GPS) and mobile communication technology have greatly facilitated the development of shared ride-sourcing services. By allowing one vehicle to serve more than one travelers simultaneously, the shared ride-sourcing services empowers more riders being served by utilizing the limited supply. Numerous benefits are provided by the shared ride-sourcing services, including the reduced trip fares for travelers, less traffic congestion, and lower greenhouse gas emissions for environmental protection. Based on whether the drivers have their own travel purposes, shared ride-sourcing services can be mainly categorized into two types: ride-pooling and ride-sharing. The former refers to situation where pooling passengers are served by dedicated drivers affiliated with ride-sourcing platforms, while the serving drivers in the latter mode have their own travel plans. Both concepts have wide applications in real life and will be investigated in this thesis. Regarding the ride-pooling service, this thesis initially examines an aggregate ride-sourcing market, where a platform provides both pooling and non-pooling services simultaneously. A mathematical model is developed to capture the complex interactions among riders’ mode choices between pooling and non-pooling, the waiting times of pooling and non-pooling users, the pairing probability, the expected detour time of pooling users, the matching rates of cruising and halfly occupied vehicles with di↵erent types of riders, and the numbers of cruising, halfly occupied and fully occupied vehicles at each instant. Based on the established model, the impacts of ride-pooling services on the total drivers’ income, the platform profit, the utilities of riders, and the social welfare are investigated both theoretically and numerically. Furthermore, this thesis attempts to optimize the pricing strategy of the mixed market, considering the temporal and spatial dynamics of the ride-sourcing market, by applying a reinforcement learning framework. For a ride-sharing service, this thesis establishes a user equilibrium model under a dynamic ride-sharing mode, where no traveler can enhance their matching probabilities by unilaterally changing their routes. Unlike existing works, which assume driver and rider demands are known before the planning horizon, dynamic ride-sharing allows driver and rider requests to occur dynamically and be matched in real time. To disentangle the complex interactions among Origin-Destination (OD) pairs, di↵erent waiting-states and seeking-states are constructed for each OD pair, and the pairing relationships between these states are well-defined. With these pairing relationships among states, a system of nonlinear equations is established to describe the user equilibrium under a dynamic ride-sharing system. By constructing a sequence of fixed-point problems, the existence of an equilibrium solution is guaranteed. An alternating optimization algorithm is then devised to solve this problem, incorporating dynamic programming to find the feasible route with the maximum number of compatible riders for each driver-OD pair.

Date of Award	2024
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Hai YANG (Supervisor) & Jin QI (Supervisor)