An effective memetic algorithm for the optimum multiuser detection problem

This work presents a new two-phase, memetic algorithm to solve the maximum-likelihood multiuser detection problem. The algorithm uses a standard evolutionary programming (EP) for the base level search, which makes a quick decision to direct the search towards the optimal region (phase-1). The local optimization by k-opt heuristic local search is then employed to do fine tuning (phase-2). In order to validate the effectiveness of the proposed method, the algorithm is applied to 10- and 20-user synchronous direct-sequence code-division multiple-access (DS-CDMA) systems. The results clearly show that the algorithm not only improves the solution quality but also makes the EP more efficient in terms of number of iterations to reach the global optimum.