An annealing approach to multi-facility location problems in Euclidean space

This paper concerns multi-facility location problems in Euclidean space, with known number of facilities, known and deterministic user demand and no costs of interaction between facilities. The authors consider three types of locations, namely medians, barycentres and centres. Iterative approximation algorithms alternate allocation and location steps until they converge to locally optimal solutions. The authors show how to make such algorithms more robust by allowing fractional allocation in combination with annealing techniques. Some experimental results on problems in R² are reported.