AI algorithms at the LPP Distribution Centre
10 Jul 2020 - Logistics, Case study
Thanks to the machine learning mechanism applied at the Fulfillment Center of the Polish clothing producer, the order-picking route got shorter and the dispatch of e-commerce orders was significantly accelerated. See how AI algorithms can efficiently solve the so-called travelling salesman problem (TSP).
LPP, the owner of Reserved, Cropp, House, Mohito and Sinsay brands, is one of e-commerce market leaders both in Poland and in Europe, handling around 11 million orders per year. Dynamic growth in online sales, fuelled also by the Covid-19 pandemics and lockdown, has made it increasingly difficult for businesses to meet the ever-growing demand from consumers. Last year, the share of LPP’s online sales reached 12%, but the pandemics caused a major shift between the traditional and online sales channels.
We have noted a 4-fold increase in the number of clients interested in online sales. As a result, it was necessary to optimise our logistics and IT systems to meet the demand without compromising the product quality and the lead time of online orders, explains Jacek Kujawa, LPP Vice President. This is why we decided to rely on our long-standing business partner, PSI Polska, to roll out the solution based on artificial intelligence algorithms which will significantly enhance the online orders processing efficiency.
With the new warehouse management system (WMS) under way, LPP can guarantee quick and uninterrupted delivery of online orders to clients.
The main task of the implemented algorithm is to efficiently solve the so-called travelling salesman problem (TSP), said Jerzy Danisz, PSIwms Standard Development Director at PSI. It boils down to finding the shortest possible route connecting a number of locations on the map. In the case of a warehouse, the system will find an optimum route connecting a dozen or so order-picking locations. At first glance it seems to be an easy task, but in fact it has been a major problem mathematicians have been trying to solve for years, concludes Jerzy Danisz.
To approach that problem, PSI Polska decided to develop a solution based on AI algorithms. A machine learning mechanism, and more specifically neural networks based on CNN units (Convolutional Neural Networks), enables to generate a list of products necessary to pick the order in real time, find the shortest order-picking route, and suggest the optimum use of resources, including warehouse equipment, fork-lift trucks and the working time of warehouse personnel. This is possible with the so-called genetic algorithm, which applies evolutionary operators (such as cross-overs and mutations) to create an optimum order-picking list.
The first tests confirmed that the new tool shortened order-picking routes by 30%, thereby significantly increasing the efficiency of order-picking processes at the Distribution Centre by as much as 11%, explains Sylwester Dmytriwski, e-commerce fulfillment general manager of LPP. The algorithm not only allows to shorten the order-picking route, but the mechanism also learns how to use the information thanks to an artificial intelligence module. It learns new data relating to the warehouse, orders and available workforce, and uses that data to group the products and facilitate the entire process, adds Sylwester Dmytriwski. Thanks to the permanent link to other warehouse resources management modules, the new system can update the information in real time and adjust generated products lists accordingly.
The order-picking route optimisation tool is part of the larger Warehouse Intelligence project based on the artificial intelligence concept, which has been recognised by the expert team at the National Centre for Research and Development and recommended for project co-financing. The new tool is to become fully operational in 2022.
The roll-out will be first launched in one of the warehouses handling LPP’s e-commerce processes. We are very happy to see that our successful business partnership between PSI and LPP and the application of agile implementation methods enabled us to efficiently roll out an innovative solution which delivered satisfactory results already in the first phase of the project, concludes Arkadiusz Niemira, President of the Board of PSI Polska.
Read also:
