Integrated task scheduling and personnel rostering of airports ground staff: A case study

• Published in: Expert systems with applications 2024-03, Vol.238, p.121953, Article 121953
• Main Authors: Cappanera, Paola, Di Gangi, Leonardo, Lapucci, Matteo, Pellegrini, Giulia, Roma, Marco, Schoen, Fabio, Sortino, Alessio
• Format: Article
• Language: English
• Subjects: Airport personnel scheduling; Integer linear programming; Integrated TSPR; Staff rostering; Task scheduling
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2023.121953

In dealing with personnel management in companies, two fundamental planning issues have to be handled: staff rostering and activities assignment. These two problems have historically been treated separately, in a sequential way; however, it is evident how strongly they are tied to each other and that solution quality inevitably drops if this connection is not taken into proper account. For this reason, also taking advantage of the massive recent advances in software and hardware technologies, the integrated task scheduling and personnel rostering problem (TSPR) has been formalized along with suitable algorithmic approaches to tackle both planning stages altogether. In this paper, we describe how the peculiar and complex case of airport ground staff was handled in a scenario defined by real-world data from a large airport in Italy. Specifically, we show that the problem can be cast into a mixed integer linear programming model. We then show that, to make the problem computationally tractable, the introduction within the model of a set of suitable valid inequalities is crucial. Indeed, as opposed to the base model from the literature, the novel, improved formulation allowed to effectively obtain near-optimal solutions in reasonable time even for the considered large-scale and real-world scenario. •The problem of simultaneously deciding shifts and duties in airports is addressed.•The resulting Integrated TSPR problem can be modeled as a MIP program.•The introduction of valid inequalities into the model is computationally beneficial.•10 large real-world instances of the problem have been solved.