Hierarchical Reinforcement Learning And Game-Theoretic Model For Stochastic Last-Mile Delivery With Crowdshipping

Abstract

Dynamic crowdshipping vehicle routing problem can be challenging as the dynamically arrived crowd drivers, orders, and the dynamic capacity and time availability of each driver. In the ecommerce context, it must also include the presence of company inhouse drivers – as they are the main source of delivery in e-commerce. The thesis applies a hierarchical reinforcement learning method that combines upper-level agent for balancing the opportunities and risks brought by delayed batch-matching, Nash game for classifying crowdsource and company orders, and lower-level agents for route planning. The model secures a 10% improvement versus novel solutions, which indicate improvements and higher chance for further development