A Learning Approach to the School Bus Routing Problem

In this paper, we introduce a solution to the School Bus Routing problem (SBRP), using a reinforcement learning technique that we previously applied in the domain of transportation logistics. This approach consists of bundling transportation requests between several locations in order to construct combinations of items in a cost-efficient way. We investigate how the combination of reinforcement learning and our novel bundling algorithm can be used to increase the efficiency in logistics and how it can be applied to other domains. In particular, we discuss how the SBRP can be transformed to resemble a problem in transportation logistics and thus solve the SBRP using our combined technique. We obtain results comparable to those presented in literature, namely from a cost minimization approach that is specifically tailored to the SBRP. We conclude that our reinforcement learning and bundling algorithms are flexible enough to be applied in different domains and offer significant reductions in the cost of the stakeholders.

Jaar van publicatie:2011

Trefwoorden:School Bus, Routing, Reinforcement Learning, Artificial Intelligence