Balancing service quality and riders’ welfare in on-demand delivery order dispatching: an integrated rolling horizon and differentiated incentive approach

As on-demand food delivery plays an important role in daily service, it also poses significant challenges in developing real-time optimization solutions. Overwhelming customer orders require real-time rider dispatch and route scheduling dynamically, while meal preparation delays disrupt the scheduled routing plan of riders. The resulting delivery lateness incurs penalties to riders. To balance delivery service quality and riders’ welfare, this paper proposes an order exchange mechanism which enables riders to return the delayed orders back to the platform. The platform then identifies potential riders willing to undertake the returned orders with differentiated incentives. Four mixed integer linear programming (MILP) models are formulated to capture the sequential interactive decisions of the platform and riders. A dynamic order returning and dispatching method is designed and embedded with the rolling horizon approach, managing both the returned orders and the newly arriving orders adaptively. Meanwhile, aimed at balancing riders’ individual welfare, we introduce an equity metric and establish a bi-objective order dispatching model, subsequently employing the ε -constraint approach with a linearization technique. A tailored Artificial Bee Colony (ABC) algorithm solves instances based on a new town in Shanghai Megacity, China. Numerical experiments demonstrate that the order exchange mechanism enhances scheduling flexibility and ensures welfare equity. Interestingly, incorporating equity considerations reveals that generous pricing strategies paradoxically improve the overall system efficiency and ultimately reduce the total operational costs. Policy recommendations based on these findings are provided for on-demand delivery service platforms to strategically manage their riders and delivery orders.