Capacity Regions and Optimal Power Allocation for Groupwise Multiuser Detection

Abstract

In this paper, optimal power allocation and capacity regions are derived for GSIC (groupwise successive interference cancellation) systems operating in multipath fading channels, under imperfect channel estimation conditions. It is shown that the impact of channel estimation errors on the system capacity is two-fold: it affects the receivers' performance within a group of users, as well as the cancellation performance (through cancellation errors). An iterative power allocation algorithm is derived, based on which it can be shown that the total required received power is minimized when the groups are ordered according to their cancellation errors, and the first detected group has the smallest cancellation error. Performace/complexity tradeoff issues are also discussed by directly comparing the system capacity for different implementations: GSIC with linear minimum-mean-square error (LMMSE) receivers within the detection groups, GSIC with matched filter receivers, multicode LMMSE systems, and simple all matched filter receivers systems.