QRP07-4: Generalized Selection Multiuser Scheduling with Channel Estimation Errors

We propose and study a generalized selection multiuser scheduling (GSMuS) scheme which ranks a total of L users awaiting transmission by their signal-to-noise- ratios (SNRs) and selects N_c users with the largest SNRs for parallel channel access. We derive the error and outage probabilities, the individual and sum rates of the proposed GSMuS scheme with adaptive modulation. The effect of imperfect channel estimation (ICE) on the sum and individual rates is analytically evaluated. Numerical results show that as the average channel SNR (ASNR) per user increases, assigning more users for parallel transmission (for a fixed total transmission power) will significantly improve the sum throughput than scheduling one best user. Under the block static fading assumption, as the ASNR increases the effect of ICE on the spectral efficiency tends to decrease, and the performance of GSMuS with ICE approaches that with perfect channel state information (CSI).