This thesis deals with optimal receiver design for pilot-assisted communication systems with imperfectly available channel state information (CSI). In conventional receiver design the channel is assumed to be perfectly known. The maximum likelihood (ML) decision metric is then derived under this assumption. But acquiring perfect knowledge of the channel poses a fundamental problem for the receiver.
In practice, the receiver performs channel estimation using techniques like least squares (LS) or minimum mean square error (MMSE) estimation.
This is accomplished by sending pilots, which are perfectly known by the receiver. Due to the limited number of pilots, the channel estimation is imperfect. The so called mismatched receiver replaces the true channel by its noisy estimate in the metric originally designed for a perfectly known channel. The resulting mismatch leads to performance degradation in terms of bit error rate (BER). In this thesis, we pursue a more advanced approach to designing a receiver by utilizing statistics of the channel and its estimation error to derive a so-called modified ML metric. The metric obtained by this method is better suited to the presence of channel estimation errors. This concept is applied in deriving a modified receiver for an iterative system architecture based on the bit-interleaved coded modulation scheme with iterative decoding (BICM-ID). Numerical simulations using an i.i.d Rayleigh block fading channel model show the superior performance of the modified receiver in terms of BER.
We further extend the idea of utilizing the channel statistics to correlated channel models and derive an optimum maximum likelihood metric for a non-iterative system architecture. The resulting optimum receiver performs sequence detection without prior channel estimation, because the received pilots are directly incorporated into the metric.
We also provide low-complexity implementation of the optimum metric.
Numerical simulations based on orthogonal frequency division multiplexing (OFDM) and autoregressive (AR) channel models show that the optimum receiver outperforms the mismatched receiver in terms of BER.
The optimum receiver is further observed to be less sensitive to the number of pilots used.