In wireless communications, having knowledge of the channel state at the transmitter side is useful to adapt the transmission scheme. To build such knowledge, one way is to predict future state of the channel from past measurements. These measurements are actually performed at the receiver side, and have therefore to be fed back to the transmitter side. To limit the feedback load, we will consider quantization on a few bits. This paper presents a MIMO channel predictor, derived from existing SISO ESPRIT-based techniques. Considering a 4 x 4 MIMO channel operating in a microcell environment with a sampling rate of 4 samples per wavelength, 17 channel observations, quantized with 2,176 bits, are sufficient to initialize the MIMO channel predictor. The complexity of the predictor remains acceptable, and its computational needs are limited. The channel predictor can also be used to predict rank of MIMO channel matrices, since knowing the rank of the channel can be sufficient for most of the transmit schemes. The strategy is to predict the channel matrix first, and then to assess its rank.
|journal||the 4th Management Committee Meeting of COST2100 - Wroclaw (Poland), 6-8 February 2008|
|Etat de la publication||Publié - 2008|