Using aggregation to reduce response time variability in cyclic fair sequences

Fair sequences are useful in a variety of applications, including manufacturing and computer systems. This paper considers the generation of cyclic fair sequences for a given set of products, each of which must be produced multiple times in each cycle. The objective is to create a sequence so that, for each product, the variability of the time between consecutive completions is minimized. Because minimizing response time variability is known to be NP-hard and the performance of existing heuristics is poor for certain classes of problems, we present an aggregation approach that combines products with the same demand, creates a sequence for the aggregated instance, and then disaggregates this solution into a feasible sequence for the original instance. Computational experiments show that using aggregation can reduce response time variability dramatically and also reduces computational effort.