Multi-user detection in alpha stable noise

We study several existing and new techniques for robust multiuser detection in the presence of alpha stable ambient noise. These methods are based on minimizing a certain cost function (e.g., the Huber penalty function) over a discrete set of candidate user bit vectors. The candidate vectors are either obtained by quantizing the continuous minimizers or based on the slowest-descent approach. Simulation results show that the new techniques offer substantial performance and convergence rate improvement over the recently proposed robust multiuser detectors, with little attendant increase in computational complexity. Furthermore, performance sensitivity on the parameter alpha of the noise distribution is low.