Using a genetic algorithm for communication link partitioning

The paper addresses two instances of link partitioning problems in communication networks: dedicated and shared partition allocation problems. These problems belong to the class of nonlinear nondifferentiable integer optimization problems which are difficult for conventional nonlinear integer programming methods to find global optima. We use a genetic algorithm for solving these problems. Possible partitions of a communication link are represented as chromosomes to which genetic operators are repeatedly applied to find better solutions. Comparative experimental results show that the genetic method outperforms uniform partitioning and a conventional heuristic method for a wide range of offered load levels in multiclass calls.