Reducing unproductive walking times is an important planning problem in modern e-commerce picker-to-parts warehouses. In such facilities, the worker traverses the warehouse to collect all items of an order. Items of the same product are hereby stored at different positions within the warehouse, such that only a subset of storage positions must be visited. The so-called picker routing problem in scattered storage warehouses aims at a shortest tour through the warehouse, which allows the picker to collect a sufficient number of items per product before returning to a decentralized depot system.

The aim of the thesis is to develop a reinforcement learning approach that optimizes the picker routing problem in mid-sized scattered storage warehouses. The famous taxi environment can be used as a starting point of the thesis.

Interested in this topic? Please use the application form on our website!