Résumé
In the last decades, e-commerce activities grew up significantly throughout the world leading to an extremely competitive environment with innovative services. The implementation of these services requires a good management of the supply chain, more precisely the storage and order picking policies inside the warehouses. These policies must be adapted to a new demand pattern characterized by a large number of orders to be prepared daily, each composed of few items. Moreover, the orders come with deadlines, and respecting these deadlines is critical to ensure high service quality. Most warehouses still follow classic picker-to-part systems in which order pickers start from a central depot, walk through the warehouse aisles to collect customer orders contained in picklists and bring them back to the assembly unit. In order to adapt to the demand pattern of e-commerce, new storage and picking techniques are implemented in these warehouses at the tactical and operational levels. The general purpose of this thesis is to propose models and methods that optimize the order picking process in warehouses using these techniques and to study their impact.
In the first part of this thesis, we study the impact of an emerging practice in e-commerce warehouses, i.e., the possibility of handling an order by several pickers (splitting an order). To this end, we assume a set of orders with preparation deadlines that must be handled by a set of pickers equipped with capacitated carts. We generalize the integrated orders batching, batch scheduling, and picker routing problem by allowing splitting an order. To solve the problem, we propose a route-first schedule-second heuristic. In the routing phase, the heuristic divides the orders into clusters and designs the tours that retrieve the items of each cluster using a split-based procedure. In the scheduling phase, the constructed tours are assigned to pickers (and sequenced) using constraint programming. On a publicly available benchmark, we show that splitting the orders reduces the picking time by 30 % on average.
In the second part of the thesis, we study the picker routing problem (PRP) (the problem consists in sequencing the storage locations that must be visited in a single tour such that the total travel distance is minimized) in a new type of warehouses, called mixed-shelves warehouses. Unlike conventional warehouses where all the inventory associated with an item is available on a single shelf in the warehouse, mixed-shelves warehouses break up the total inventory into smaller units that are assigned to different shelves throughout the warehouse. To solve the problem, we propose a logic-based Benders decomposition method in which the storage locations from where to retrieve the items are selected in the master problem and the tour that visits the selected locations is determined in the sub-problem. We design a tailored optimality cut and introduce a set of algorithmic enhancements that significantly improve the method. Finally, a comparison with an adaptation of an exact method, originally proposed for the PRP, to the context of mixed-shelves warehouses demonstrates the superiority of our method.
In the last part of this thesis, we adapt the method proposed in the second chapter to solve a problem that combines two techniques, namely mixed shelves storage and zoning (dividing the picking area into disjointed zones with a picker operating in each zone). In our problem, a wave of orders must be prepared by collecting all items of the orders in the next tour of each picker. The objective is to select from where to retrieve each item of the orders and to define the tour associated to each picker such that the makespan is minimized (the longest tour among the pickers). We demonstrate in the experimental part that the implementation of a zoning strategy in mixed-shelves warehouses reduces significantly the makespan compared to zoning in conventional warehouses.
Source: http://www.theses.fr/s238830
.