Order Fulfillment Optimization in Automated Warehouses

Abstract

In warehousing, order batching is one of the most popular strategies for optimizing order fulfillment as it groups similar orders into the same batch to optimize picking. The order similarity can be determined based on item locations, availability, and order compositions. The objectives include minimizing travel time, maximizing the number of picked items, and maximizing simultaneous multi-order processing. In this thesis, we study the order fulfillment problem in automated warehouses and propose an order fulfillment heuristic method that to minimize the number of required pick-up sequences to fulfill given order lists by integrating various independent order fulfillment techniques. Three independent algorithms are modified and integrated: (1) FP-Growth-based Association Rule Mining, (2) Order Batching using Similarities Between Orders, and 3) A Hybrid of Public and Personal Item Storage. The resulting heuristic approach is capable of finding optimal solutions when compared to exact results based on Integer Programming. Additionally, a custom-built Python simulation platform is created and run to prove the scalability of the devised algorithm. The Python simulation platform has been further developed into an ROS- and Gazebo-communicable simulation platform for more visualized and intuitive simulation results. Based on the simulation results involving 2000 orders and 1000 items, the algorithm reduced the total number of required pick-up sequences by approximately 50% in comparison to traditional First-Come, First-Served approach.