Parallel evolutionary algorithms in telecommunications: two case studies

Abstract

Sequential and parallel evolutionary algorithms (EAs) are developed and evaluated on two hard optimisation problems arising in the field of Telecommunications: designing error correcting codes, and finding optimal placements for antennas in radio networks. Different EA models (generational, steadystate and cellular) are compared on these two problems, both in sequential and parallel versions. We conclude that the cellular EA is a very effective technique, consistently finding the optimum, although it is slower than a steady-state EA. A distributed steady-state EA is shown to be the best approach, achieving the same success rate than the cellular EA in much lower time. Furthermore, it is shown that linear speedups are possible when using separate processors.