Channel Prediction Aided Multiuser Transmission in SDMA

Transmit preprocessing employed at the basestation (BS) has been proposed for simplifying the design of the mobile receiver. Provided that the channel impulse response (CIR) of all the BS to mobile station (MS) links is known in advance-even before the signal's transmission-it is plausible that the different users' signals may be differentiated with the aid of their unique, user-specific downlink CIRs. Naturally, this non-causal CIR knowledge is unavailable in practice. Hence a natural design option is to estimate the CIRs at the receiver after the BS's signal was received and convey it using side-information to the BS for its future use. Naturally, the resultant CIR has to be quantized before its transmission. In addition to this quantization error, it also becomes outdated and both imperfections result in an erosion of the achievable transmit preprocessing gain expressed in terms of either the attainable transmit power reduction or the number of users that may be supported. Another attractive design option is to avoid the CIR-signalling latency by invoking the previously received CIRs for predicting their future evolution using CIR-tap prediction. In this paper, Kalman filtering aided CIR prediction is combined with both minimum mean square error (MMSE) and zero forcing based preprocessing. Our simulation results show that the proposed scheme is capable of attaining 6.5 dB gain at a BER of 10^-2, when using 6 antennas.