Cooperative Genetic Algorithm for Optimization Problems in Distributed Computer Systems

In the proposed algorithm, several single population genetic algorithms with different cross-over and mutation parameters are run as a set of processes that cooperate periodically and exchange information to solve the problem efficiently. The algorithm is less stochastic than the standard genetic algorithm and a distributed implementation is appropriate for application to large scale problems. In particular, we apply it to the static task assignment problem and suggest modifications to solve other optimization problems in distributed computer systems. Preliminary experiments with fairly large-sized problems of allocating 50 tasks among 16 processors indicate that the cooperative algorithm implemented on a network of workstations quickly finds better solutions than those obtained by a standard genetic algorithm. To conclusively show that better solutions are obtained, extensive experiments have to be performed. A distributed implementation of the algorithm is highly suited for such expe..