Fair and Stable Allocation in On-Demand Delivery Services for Meals and Groceries

Most existing studies in the shared mobility literature address the request-vehicle assignment problem with a globally optimal goal, with only some consideration to the parties involved. This study deviates from the norm and employs a decentralized approach called stable and fair matching algorithm (SFMA) for the two-sided matching problem between requests and vehicles for on-demand delivery (ODD) of meals and groceries. The SFMA matching pairs are stable and fair such that no pair of requests and drivers prefer to change the match. With meal preparation and grocery packaging time considered in simulation, a case study in the metropolitan region of Austin, Texas is conducted with POLARIS, a large-scale agent-based mesoscopic traffic simulator, to illustrate the matching performance of SFMA. The delivery services are provided by operators closely resembling transportation network companies (TNCs) in the simulation. Results are compared to the existing default heuristic strategy (DHS) to demonstrate the SFMA benefits in terms of the average wait time, matching rate, vehicle usage rate, empty vehicle miles travelled (eVMT), and the average profit of vehicles. Several scenarios are investigated to assess the impacts of fleet size on performance of SFMA. Compared to DHS, SFMA improves the matching rate and profits earned per vehicle due to the preference consideration of TNC drivers while the resultant average wait times and eVMT increases slightly.