|Ph.D Student||Levy Nathan|
|Subject||Information Theoretic Aspects of Cellular Communication|
Models with Joint Multiple-Cell Processing
|Department||Department of Electrical and Computer Engineering||Supervisors||PROF. Shlomo Shamai )Shitz(|
|PROF. Ofer Zeitouni|
In this work, we study multiple cell-site processing from an information theoretical point of view as a way to enhance the performances in Wyner-type cellular communication settings. In a first part, we focus on models featuring random fading. Due to the particular sparsity of the channel transfer matrix, we cannot use the classical tools of random matrix theory. We use the fact that the coefficients of the channel transfer matrix are non-zero only in a narrow band around the diagonal, which allows us to use techniques coming from the theory of random Schrödinger operators.
We derive the per-cell sum-rate capacity as a function of Lyapunov exponents of a sequence of random matrices and in several special cases we derive closed formulas in the high-SNR regime. We also study the fluctuations of the per-cell sum-rate capacity in the non-ergodic regime and provide results of the type of central limit theorem and large deviations. Finally, using information theoretical tools, we derive bounds on the asymptotic per-cell sum-rate capacity, which allow us to study the influence of different key parameters of the system.
In a second part, we relax the multiple cell-site processing hypothesis and we study the performances of clustered local decoding in several settings. We first study the distribution of the achievable rate and the outage probability in a users' activity problem. Then, we consider several two-dimensional settings and a one-dimensional soft-handoff setting and derive several lower and upper bounds on the achievable rate by rate splitting and genie-aided bounds, originally employed in the framework of the interference channel. Finally, we derive the high-SNR multiplexing gain for two communication settings featuring both clustered local decoding and cognitive users.