# Minimising the walking distance of pickers on a unidirectional picking line

Speaker

## Abstract

An order picking system in a retail distribution centre (DC) operated by PEP Stores Ltd. (PEP), located in South Africa forms the focus of the presentation. PEP is mainly trading in apparel and the DC services about 1800 retail stores scattered over the whole of the country.

PEP employs central planners. Each planner manages a certain type of product (for example, boys’ short pants) for all the stores. These planners thus decide on how much of each product are send to each store. Thus a central planning office releases batches of stock keeping units (SKUs) that differ only in size. These batches are called distributions (DBNs) and are continuously released to the DC to pick and ship to the stores. A DBN thus contains one or more SKUs together with the instructions of how many of each of the SKUs in that DBN should be sent out to each store receiving that SKU. Each DBN also comes with a deadline in the form of an out-of-DC date. SKUs in a DBN must be kept together throughout the order picking process to ensure that all the sizes of a product arrives at the same time at the stores.

The order picking system consists of parallel unidirectional picking lines. These picking lines resembles carousel picking systems. Order picking happens in waves. A wave of picking implies that a set of DBNs is assigned (and physically taken) to a specific picking line. A team of pickers then pick all the SKUs on this picking line for all the stores. Once all the SKUs are picked (and leftover stock, if any, are removed) from the picking line, the picking line is populated with a new set of DBNs to start a new wave of picking.

In this picking process three optimisation problems/questions arise, namely:

1. What is the optimal set of SKUs to be grouped together on a picking line during a wave of picking?
2. What is the optimal placement of the selected SKUs (in 1) on the picking line?
3. What is the optimal sequence in which all the orders should be picked for the combination and placement of SKUs (determined in 1 and 2) to minimise the total travel distance of pickers?

This talk will highlight the importance of all three problems, but the focus is on solving the third optimisation problem. Although the sequencing problem can be formulated as an IP it cannot be solved in a reasonable timeframe. Different greedy heuristics, metaheuristics and a tight lower bound are thus introduced, tested and compared with real historical solutions used by PEP. Both the heuristics and metaheuristics perform well relative to the lower bound. The recommendation is thus that the best fast greedy heuristic should be used, because it helps in speeding up the solution times of the first two questions.