Università degli studi di Modena e Reggio Emilia

In healthcare supply chain, centralised warehouses are used to store large amounts of products close to hospitals and pharmacies in order to avoid shortages and reduce storage costs. To reach these objectives, the warehouses need to have efficient order retrieval and dispatch procedures, as well as a storage allocation policy able to guarantee the safe keeping of items. Considering this scenario, we present a Storage Location Assignment Problem with Product-Cell Incompatibility and Isolation Constraints, that models the targets and restrictions of a storage policy in a pharmaceutical product warehouse. In this problem, we aim to minimise the total distance travelled by the order pickers to recover all products required in a set of orders. We propose an Iterated Local Search algorithm to solve the problem, and present numerical experiments based on simulated data. The results show a relevant improvement with respect to a greedy full turnover procedure commonly adopted in real life operations.

We study a problem from a real-world application, in which a daily set of orders must be processed following two stages, consisting of preparing perishable products on benches and allocating them to conveyors to be packed in disposable trays. Daily decisions must be made regarding the number and start time of working shifts, the number of workers and their allocation to machines, and the scheduling of orders in a two-stage flexible flow shop environment. The flow shop environment of the studied problem is common in many industries of perishable products, making the problem very general. The problem involves a number of operational constraints, and three objective functions that are minimised in a lexicographic way. To solve the problem, we implement a constructive heuristic and embed it within three metaheuristics: a Random multi-start algorithm (MR), a Biased random key genetic algorithm (BRKGA), and a Variable neighbourhood search (VNS) based one. We perform computational experiments over a set of realistic instances, and present a lower bound obtained from a constraint programming model for the scheduling counterpart. The results of the experiments show that the BRKGA is the most effective in practice for the integrated problem of workforce allocation and scheduling.

The management of queues is a complex problem, and it requires special attention in dynamic environments where information changes over time. This work focuses on an outpatient facility system where patients are attended by identical parallel servers offering different services. Each patient requires service and expects to receive it within a given target time, after which, a tardiness is created. The objective of the problem is to minimize the total tardiness while defining which services each server will offer during the working hours. The arrival of patients is dynamic, and the server’s configurations of services can be updated from time to time. To solve the problem, we propose a local search-based heuristic that locally assigns a configuration to each server based on the improvement reached in terms of total tardiness. The heuristic is tested on realistic instances, considering different settings, showing its superiority over the solution currently implemented on the facility system.

We address a problem from a meat company, in which orders are produced in two stages, consisting of preparing meats on benches and allocating them to conveyors to be packed in disposable trays. In an environment where machines are unrelated, the company has to take daily decisions on the number and start time of working periods, the number of workers and their allocation to machines, and the scheduling of activities to satisfy the required orders. The objective of the problem is to minimize, in a lexicographic way, the number of unscheduled activities, the weighted tardiness, and the total production cost. To solve the problem, we propose a multi-start random constructive heuristic, which tests different combinations of number of workers in the machines and for each combination produces many different schedules of the orders. The results of our computational experiments over realistic instances show that the heuristic is effective and can support the company on its daily decisions.