Introduction:

The capacitated vertex p-center problem is a location problem that consists of placing p facilities and assigning customers to each of these facilities so as to minimize the maximum distance between any customer and its assigned facility, subject to demand capacity constraints for each facility.

Method:

In this work, a metaheuristic for this location problem is presented. It integrates several components such as a greedy randomized construction with an adaptive probabilistic sampling scheme and an iterated greedy local search with variable neighborhood descent.

Results:

Empirical evidence over a widely used set of benchmark instances on location literature, reveals the positive impact of each of the developed components and the quality of the solutions delivered by the heuristic when compared with existing methods. For instance, the proposed heuristic was able to find feasible solutions to all but two instances, while the best of the existing methods failed in 18 of these instances.

Conclusion:

It is found empirically that the proposed heuristic outperforms the best existing heuristic for this problem in terms of solution quality, running time, and reliability on finding feasible solutions for hard instances.