MIMO signal design

Wireless communication systems employing multiple antennas have many attractive properties such as high spectral efficiency and research in this field has seen an explosive development in the last decade; the introduction of space-time coding being one driving source \cite{Tarokh:98}. A limiting factor for the efficiency is the accuracy of the information regarding the channel characteristics available at both transmitter and receiver ends. Such information can be available in the form of statistics and/or estimates based on measurements. Estimation can be performed in two ways: blind estimation where the transmitted signal is unknown, or using a known training (or pilot) signal. The training signal can also be superimposed on transmit data.

Optimal design of the training signal is a way to improve the channel estimate in training based methods. Maintaining efficiency requires these signals to be short. Despite the fact that there is by now a significant body of literature on this problem for various scenarios
we have identified several directions which we believe are of importance for the further developments of this field:

  1. Performance oriented training. Almost all contributions base the training signal design on a lumped measure of the quality of the channel estimate that only indirectly accounts for the quality of the reconstructed signal at the receiver side. A frequently used measure is the mean-squared error (MSE) of the channel matrix estimate. It would be more appropriate to consider the impact the training signal has on the quality of the reconstructed signal, e.g. measured by its MSE or by the bit error rate.

  2. Convexification of optimal training signal design programs. For certain communication scenarios it has been shown that there exist explicit solutions to optimal training signal design problems. However, existing cases are restricted to certain structures, e.g. flat fading channels subject to noise possessing a certain Kronecker structure, and to certain performance measures, such as the MSE of the estimated channel matrix. We believe that significant progress can be achieved by asymptotic (in the training signal length) approximation of the problem and using recent advances in convex optimization and experiment design for dynamical systems. We believe that this approach also will allow to develop simplified design techniques for many common scenarios.

  3. Benefit analysis. With much of current efforts having been devoted to algorithmic developments, we believe that it is now important to analyse for which scenarios training signal design really makes a difference. Also here we believe that asymptotic approximation can be useful.

Project team

Division of Decision and Control Systems KTH

Division of Signal Processing KTH

Project funding and duration

This project is funded by the Swedish Research Council with duration 2011–2013.

Håkan Hjalmarsson
Professor of Signal Processing

My research interests cover system identification, process modeling and control, and communication network