A Reinforcement Learning Approach for Crowdshipping in Food Delivery: Role of Pricing Decisions

The recent significant increase in the online food delivery market both in the US and many other countries and the emergence of transformative mobility services have brought both challenges and opportunities. While their evolution has provided consumers with a variety of options, existing services lack transparency in how they operate and their fee structure. Moreover, their long-term impact on human health, the environment, and social welfare is either unknown or unexplored. This study formulated and solved a simulation and optimization of a pricing model based on reinforcement learning techniques to establish dynamic, zone-based pricing for online food delivery services. A sample of restaurant or other food outlet dining experiences in San Francisco was used as a case study to test the model’s performance. The designed pricing method outperformed the alternative static and myopic pricing strategies. The results also reveal the importance of platform providers’ decisions regarding total profit and deliveries. Optimal surge pricing coupled with the efficient distribution of drivers in delivery regions is an important means of improving a service’s impact on the environment and social welfare. Microanalysis has concluded that the change in general cost per customer relies heavily on the value they place on the time they save using online food delivery services as an alternative to dining out. However, the elimination of active transport time among those who used to bike or walk to access food outlets for their meals negatively impacts their health in the long term. This could potentially be mitigated by ordering healthy food or engaging in other forms of physical activity during the time saved ordering food online