| In wireless communications, the wireless channels may exhibit frequency-selectivity and time-variability, and channel coefficients may have large and small scale fadings. All of these effects, together with thermal noise, severely impair the received signals. The receivers are thus designed to undo these effects to recover the transmitted signals. In practice, receivers usually utilize channel state information (CSI) that is obtained through training-based channel estimation. In the presence of channel noise, CSI can hardly be estimated perfectly. In multi-user scenario, what is even worse is that one user may be interfered by other users, thus causing the estimated CSI of one user being contaminated by that of other users. Furthermore, pilots are sometimes inadequate or unavailable, and therefore only partial CSI is known at receiver.;This dissertation is focused on designing robust receivers for a number of non-ideal scenarios. For most receivers, detection and decoding are separated as two sequential steps, that is, output of the detector is fed to a downstream decoder. The major obstacle of combining the two steps lies in the fact that detection is performed on real or complex field, whereas decoding is usually on finite field. In the recent decade, a new decoding scheme in the form of linear programming (LP) gains wide popularity since the seminal work by Feldman. The LP decoding opens the door for a new era of receiver design that integrates detection and decoding stages. Unlike the quasi-joint turbo receiver exchanging extrinsic information iteratively between detector and decoder, we propose receiver formulations that jointly detect and decode received signals in use of LP decoding constraints. We name the proposed design paradigm the code anchored robust design (CARD). The effectiveness of CARD receivers is demonstrated by numerical results that establish substantial performance gain of the proposed receivers over existing designs. In addition, the designed joint receivers can be extended for turbo processing, and can also be used in distributed multi-cell processing for improved performance. |
