Picking location metrics for order batching on a unidirectional cyclical picking line

In this paper order batching is extended to a picking system with the layout of a unidirectional cyclical picking line. The objective is to minimise the walking distance of pickers in the picking line. The setup of the picking system under consideration is related to unidirectional carousel systems. Three order-to-route closeness metrics are introduced to approximate walking distance, since the orders will be batched before the pickers are routed. All metrics are based on the picking location describing when a picker has to stop at a location to collect the items for an order. These metrics comprise a number of stops, a number of non-identical stops and a stops ratio measurement. Besides exact solution approaches, four greedy heuristics as well as six metaheuristics are applied to combine similar orders in batches. All metrics are tested using real life data of 50 sample picking lines in a distribution centre of a prominent South African retailer. The capacity of the picking device is restricted, thus the maximum batch size of two orders per batch is allowed. The best combination of metric and solution approach is identified. A regression analysis supports the idea that the introduced metrics can be used to approximate walking distance. The combination of stops ratio metric and the greedy random heuristic generate the best results in terms of minimum number of total cycles traversed as well as computational time to find the solution.