Box-constrained multiuser detection based on multiplication-free coordinate descent optimisation

Multiuser detection can often be described as a constrained quadratic optimisation problem. Many iterative techniques are available to solve this problem; however, due to high complexity they cannot be effectively implemented in real-time. We propose a novel iterative technique, the DCD (dichotomous coordinate descent) algorithm which guarantees convergence in the box-constrained quadratic optimisation problem. We investigate application of the DCD algorithm to multiuser detection and show that it provides a high detection performance and complexity lower than other known box-constrained multiuser detectors. Moreover, the proposed algorithm is multiplication-free and division-free; this makes it attractive for real-time implementation in hardware (FPGA, ASIC) or fixed point (DSP) software.