Harmony Search Based Algorithms for the Minimum Interference Frequency Assignment Problem

The Minimum Interference Frequency Assignment Problem (MI-FAP) is a particularly hard combinatorial optimization problem. It consists in the assignment of a limited number of frequencies to each transceiver of the network without or at least with a low level of interference. In this work, we present an adaptation of the Harmony Search (HS) algorithm to tackle the MI-FAP problem. The results obtained by the adaptation of the classical Harmony Search algorithm are unsatisfactory. We performed a computation testing over some data sets of various sizes picked from public available benchmarks. The experimental results show that the conventional harmony search suffers from its premature convergence and therefore gets stuck in local optima. Even when it succeeds to escape from the local optimum, it does it after a long period of time. This make the process very slow. Due to these unconvincing results, we want to improve the Harmony Search algorithm’s performances. To handle that, we propose some small changes and tricks that we bring to the original Harmony Search algorithm and a hybridization with a local search and the Opposition Based Learning (OPBL) principle. Here, we propose two strategies to improve the performances of the classical harmony search algorithm. We will show that both of them succeeds to enhance the performances of the harmony search in solving the MI-FAP. One of the proposed strategies gives as good results as those of the state of the art for some instances. Nevertheless, the method still need adjustment to be more competitive.