Staff Scheduling with Particle Swarm Optimisation and Evolution Strategies

Sub-daily Staff Scheduling for a Logistics Service Provider

“…The mutation scheme based on maximum entropy provides better results for the ES than a more traditional approach based on rounded Gaussian mutations as given in [19]. This result underlines the great importance of adapting the mutation operator to fit the characteristics of the search space as good as possible.…”

“…Repairing solutions significantly improves the quality of results, but also doubles the runtime requirements of the heuristics from roughly 25 minutes in [19] to 50 minutes here for a single run of PSO and ES. This effort, however, is acceptable as there is sufficient time available for creating the schedule.…”

“…In standard-ES mutation is performed using normally-distributed random variables so that small changes in a solution are more frequent than large changes. In [19] we developed a search operator that adheres quite closely to this classical form of mutation and produced fairly good results.…”

“…A PSO approach and evolution strategy for this application are outlined in the following section. This work builds upon previous research [19]. We now give a brief general overview of particle swarm optimization and the evolution strategy.…”

“…The suffix +R indicates the use of the repair heuristic. Results for PSO without repair are based on[19] …”