GMD: a new CDMA multiuser detection technique using an evolutionary algorithm

A new multiuser detection technique, based on the genetic algorithm, has been proposed and its performance is evaluated via simulations. The technique is the genetic implementation of optimal multiuser detection (OMD) using a new concept for the search space to process and improve the outputs of matched filters in terms of suppression of multiple access interference (MAI) in a code-division multiple-access (CDMA) system. A new matrix structure is formed from the conventional detector. The elements of this matrix, using the new search space, new fitness function, and employing a chromosome-like structure, evolve to make a near-optimal solution to nearly eliminate the MAI. Unlike neural networks, the introduced genetic engine does not need a complicated training process. When it reaches an acceptable level of performance, it has the unique capability to trigger off the genetic engine, dropping the weaker individuals and continuing with a smaller group of the fittest selected solutions. The genetic multiuser detector (GMD) outperforms the well known multiuser detection techniques such as multistage detectors in terms of bit error rate. By increasing the population size of the individuals, the GMD reaches near the single user bound with much less complexity than the OMD