A Fuzzy Crew Rostering Model Based on Crew Preferences and Seniorities considering Training Courses: A Robust Optimization Approach

Crew scheduling problem is divided into crew pairing problem (CPP) and crew rostering problem (CRP). In this paper, a rostering model is presented to assign crew to pairings in such a way that total weighted preference is maximized. Crew members declare which parings they wish to be assigned and which ones are undesirable for them. A score is calculated in the objective function if a crew member is assigned to his/her preferred pairing, and a penalty is considered if he/she is assigned to an undesirable pairing. Moreover, crew seniorities are considered in calculating total preference. In addition, the model considers standard rules and regulations as well as crew attendance at the required training courses. The model is formulated in such a way that inconsistent crew members are not assigned to a flight. Due to the uncertainty in determining of the seniority weight, this parameter is considered as fuzzy. At the end, the robust counterpart of the nominal model is developed due to the uncertainty of time away from the base (TAFB). In this research, the issue of inconsistent crew in rostering problem is considered for the first time. Moreover, a new scoring mechanism is introduced to calculate desirable and undesirable assignments in the objective function. The proposed CRP is solved using the genetic algorithm (GA), and its performance is verified in comparison with GAMS in some test problems. On average, the optimality gap in GA is only 0.5 percent. Finally, the proposed model is examined with real-world data from Air India Airline. In comparison with the previous research studies, the suggested model (scoring mechanism) reduced the number of undesirable rosters by 61.59%.

Automated summary

A crew rostering model is proposed in this study, taking into account crew preferences, undesirabilities, and seniorities. The model is solved using a genetic algorithm and compared with GAMS. Results from real-world data show that the model is effective in reducing undesirable rosters.