Steiner TSP based on aisle as a unit for order picking. (June 2022)

• Record Type: Journal Article
• Authors: Su, Yixuan; Li, Meixia; Zhu, Xi; Li, Chunfa
• Is Part Of: Computers & industrial engineering. Volume 168(2022)
• Publication Date: 2022-06
• Subjects: Warehousing -- Order picking -- Steiner TSP -- Multi-block layout -- Picker routing -- Analytical and numerical modelling -- Performance analysis; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2022.108026
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: A new Steiner graph with exible nodes based on aisle as a unit is proposed for the Steiner TSP. Different picking access modes as well as restrictions on connections for picking aisles are introduced to meet the rationality of the route. The cost of the optimal route is divided into the cost between Steiner points contained in the route and the cost of access modes for all required picking aisles. A mixed integer linear programming model based on aisle as a unit is presented and has good overall performance, especially in large scale cases. Abstract: The picker routing problem (PRP) is one of the most crucial problems in warehouse management. It is also defined as the Steiner Traveling Salesman Problem (Steiner TSP). We deal with the Steiner TSP for a low-level picker-to-parts order picking system within a multi-block warehouse, which is much more complicated to be solved compared with those in a single block layout. After a series of preprocessing, such as reducing the vertices, transforming picking aisles and limiting the connections between vertices, we construct a Steiner graph with flexible nodes based on aisle as a unit. Different picking access modes as well as restrictions on connections for picking aisles are introduced. Then, a mixed integer linear programming model for Steiner TSP based on aisle as a unit is presented. We compare the numerical computation on the benchmarks in exiting literatures. Our Steiner graph can greatly reduce the scale of vertices and arcs without losing characteristics of the original problem. And the model has good overall performance, especially in large scale cases.