Abstract

The recent proliferation of food delivery services with crowdsourced couriers (delivery workers) has introduced new problems in the operations of on-demand food delivery platforms. An increase in crowdsourced couriers has diversified delivery transportation modes such as motorcycle, bike, and walk, which results in mode-specific maximum delivery distances according to their speeds. In addition, it is necessary to manage various customer classes with specific delivery distances between restaurants and customers. For a food delivery platform with multiple customer classes and courier pools, it is essential to optimize the delivery price charged to each customer, the wage paid to couriers, and matching between them to complete the food delivery within a reasonable amount of time. Hence, we study the operations decisions for the online food delivery platform that faces price- and time-sensitive customers with heterogeneous delivery distances and earning-sensitive couriers with heterogeneous transportation speeds. We formulate this problem as a multi-class multi-pool queueing and allocation problem. We find that the objective function of this optimization problem is inherently complex with respect to the number of customer classes and courier pools, making it unsuitable for standard convex optimization techniques. Hence, we propose a new Utilization-Based Multipoint (UBM) heuristic that exploits multiple initial points to derive near-optimal results efficiently. Our numerical experiments on both generated instances and real delivery data demonstrate that the proposed heuristic algorithm significantly outperforms the benchmark for all experiments.