Author: Dalton Griner, Kathryn L. Kolsky, Joseph John Lucido, Andrew J. Veres 👨🔬
Affiliation: Mayo Clinic 🌍
Purpose: This project aimed to automate a complex and time-consuming employee scheduling process. By replacing the traditional manual method with a multi-criteria optimization-based system (MCO), the goal was to expedite scheduling, incorporate a wide range of constraints and preferences, ensure equitable task distribution, and reduce human bias. Additionally, a robustness analysis was conducted to identify critical scheduling points for better forecasting and resource allocation.
Methods: A large-scale integer programming model was developed using a constrained programming with satisfiability methods solver (CP-SAT) from the OR-Tools library. Numerous hard and soft constraints were encoded, including fairness objectives, employee preferences, and penalty minimization for constraint violations. Custom objective functions balanced these factors to generate an optimized schedule. A novel normalized robustness score (NRS) was introduced to evaluate the impact of removing any assigned employee on overall schedule feasibility, thus highlighting tasks and days most vulnerable to resource shortages.
Results: Compared to manually created schedules, the automated solution yielded a lower total penalty score, reduced scheduling time by a factor of 50, and more evenly distributed tasks across personnel (average deviation of assignments +/- 2.5 for the automated schedule compared to +/- 4.0 for the manual schedule). Employee preferences—such as project or personal development days and preferred shifts— were accommodated more effectively as observed by lower penalty scores for preference constraints. The robustness analysis pinpointed specific assignments where employee removal posed a high risk of disruption, offering critical insights for contingency planning. Furthermore, the automated method eliminated subjective biases inherent in manual scheduling.
Conclusion: These findings demonstrate that a CP-SAT-based automated scheduling system can substantially improve efficiency, fairness, and resilience. By providing equitable task assignments and highlighting vulnerable points in the schedule, the approach supports better forecasting and resource allocation while removing human bias from the process.