Dynamic Order-Driver Matching Framework Design for Crowd-Sourced Delivery Considering the Drivers’ Acceptance Uncertainties

In this paper, we consider a crowd-sourced delivery system where in-store customers, as the supplement to a professional fleet, deliver online orders while on their way home. This system operates in a dynamic and stochastic setting: both in-store customers and online orders arrive continuously throughout the day, and the in-store customers are free to decide whether to accept or reject the assigned online orders. To address the inherent matching challenges presented in this system, we propose a two-stage dynamic order-driver matching framework, which processes an order postponing algorithm integrating the concept of “Value of Time” and a formal matching and compensation decision process employing a well-designed binomial logit model to predict drivers’ order acceptance probability. Computational findings suggest that our method provides significant economic and societal advantages. Furthermore, we delve into how varying environmental factors influence the performance of our approach.