Performance bounds for group decoding over asynchronous Gaussian multiple-access channels

Group decoding receivers (GDRs) are an attractive way to improve the performance of multiuser systems with highly correlated users. These reduced complexity receivers use jointly the coding information of a subset of users and multiuser information. Performance of GDR for synchronous multiple-access channels is upper bounded by E.A. Fain and M. Varanasi (see IEEE Trans. on Commun., vol.49, no.6, p.1021-32, 2001). However, upper bounds of GDR for asynchronous multiple-access channels (where the effect of error propagation is critical), have not been reported so far. We derive an upper bound for the performance of GDR for asynchronous multiple-access channels considering error propagation effects. We show that performance analysis based exclusively on minimum distance error events is poor, even at high signal-to-noise ratios (SNRs). To improve the accuracy of analytical estimates, new bounds accounting for the contribution of non-minimum distance error events and error propagation are derived. Comparisons with values obtained from computer simulations show the excellent accuracy of the proposed bounds. Our results also reveal that performance of GDR in channels with highly correlated users is severely affected by error propagation. To mitigate this problem, we propose a new scheme based on the combination of the generalized Viterbi algorithm and the GDR. We show that the proposed receiver is able nearly to eliminate error propagation with low additional complexity.