Fundamental limits of robust interference management: from content-oblivious to content-aware wireless networks

Abstract

In this thesis we progress towards the understanding of the fundamental limits of wireless networks with partial instantaneous channel state information at the transmitter (CSIT). We first consider classical content-oblivious networks, where the edge-nodes are unaware of the kind of requested content. We study the K-user multiple-input-single-output (MISO) broadcast channel (BC), where a K-antenna transmitter serves K single-antenna users, and we characterize the optimal degrees-of-freedom (DoF) region under arbitrary CSIT levels for the users. We then study the overloaded MISO BC with two groups of CSIT qualities. We propose a transmission scheme where no CSIT codewords are superimposed on top of spatially-multiplexed codewords. We show that the developed strategy outperforms the existing schemes and achieves the entire DoF region. Next, we move from content-oblivious networks to content-aware networks, where the edge-nodes can predict the most popular content. We first consider the K-user cache-aided MISO BC, where users are equipped with a cache memory. For a symmetric setting, in terms of channel strength levels, partial channel knowledge levels and cache sizes, we characterize the sum-generalized-degrees-of-freedom (sum-GDoF) up to a constant multiplicative factor of 12. We further show that the characterized order-optimal sum-GDoF is also attained in a decentralized setting, where no coordination is required for content placement in the caches. We then study the cache-aided interference channel, where an arbitrary number of cache-equipped transmitters serve an arbitrary number of cache-equipped receivers. Transmitters communicate with receivers over two heterogenous parallel subchannels: one with perfect CSIT, and the other with no CSIT. Under the assumptions of uncoded placement and separable one-shot linear delivery over the two sub-channels, we characterize the optimal sum-DoF to within a constant multiplicative factor of 2. We extend the result to decentralized setting, and we characterize the optimal one-shot linear sum-DoF to within a factor of 3.