Finding Optimum Solutions to the Shift Minimization Personnel Task Scheduling Problem With a New Pack-based Approach

Workforce scheduling is an important practical problem whose solution quality significantly affects an organization’s finances, morale and quality of service. It follows that good algorithms for various shift generation problems are needed for efficient real-world solutions. The Shift Minimization Personnel Task Scheduling Problem (SMPTSP) is to assign all the tasks to the minimum number of employees s.t. the employees have the requisites to carry out the tasks assigned to them, and that no two tasks assigned to the same employee overlap in time. We present a new pack-based heuristic for SMPTSP. The heuristic combines a powerful pack-based local search operator with a reinforcement learning weighting scheme for the tasks. The new heuristic performs better than the previously published methods. For almost all of the known 277 benchmark instances, the heuristic finds the optimum solution within ten seconds.