Master defense by candidate Niccolò Franceschi.
The advance of MIMO techniques as a means of boosting data rate and reliability in wireless communications has challenged researchers to investigate new channel estimation methods. As MIMO multiplies the number of channel parameters, longer pilot-sequences need to be sent to attain the same accuracy as SISO. Of course, this increases the overhead resulting in a waste of channel capacity. Semi-blind channel estimators address this problem making use of both pilot-sequence and user data to enhance the quality of the estimate. Even though these methods are appealing in terms of mean squared error, they considerably raise the complexity of the receiver. This issue is even more severe if we consider that MIMO is expected to speed up the bit rate, meaning that an increasing amount of data has to be processed to produce the estimate.
The aim of this thesis is investigating low-complexity semi-blind estimation techniques, capable of improving the mean squared error and still computationally affordable. Firstly, the MIMO-LTE channel model is formulated, then we will discuss traditional pilot-only estimation and its limitations. Afterwards, the semi-blind problem is presented and expressed using two different approaches, one relying on the true discrete distribution of the data symbols and the other on a Gaussian approximation. Then, EM-based solutions are derived and compared with numerical techniques that are independent of the size of the data sequence. Finally, all these methods are tested through simulations assessing their accuracy and computational cost.
Supervisors: Lars Christensen (Renesas Mobile), Søren Christensen (Renesas Mobile) and Ole Winther (DTU Informatics)