Genetic algorithms for multiuser detection in synchronous CDMA

The optimal multiuser detector for communication systems can be characterized as an NP-hard optimization problem. The application of genetic algorithms to suboptimal multiuser detection is studied. The performance of the genetic algorithm multiuser detection is studied by computer simulations. They demonstrate that a good initial guess vector is needed to obtain a superior performance. If the decorrelating detector output is included in the set of initial guess vectors, the performance of the genetic algorithm detection is close to that of the single user system.