ZHANG Jingjing

Thesis

Interplay between caching, feedback and topology in the wireless broadcast channel

Current PHY-based communication solutions do not successfully scale in large networks, because as the number of users increases, these solutions cannot fully separate all users' signals. This problem in turn can gradually leave each user with small communication rates, and it comes at a time when wireless data traffic is expected to increase dramatically. Recently a new technology coded caching was proposed, which was based on pre-caching content from an existing library of many popular files. This promising approach naturally came with its own limitations though. With this as a starting point, the main work of this thesis on advanced wireless caching, is about exploring the deep connections between caching and fundamental primitives of wireless networks, such as feedback and topology.
 
In the first part we reveal the interesting relationship between caching and CSIT feedback, which can potentially allow for a combination that bypasses each approach's limitations. In particular, we first consider coded caching and delayed feedback, and show a powerful synergy between them, in the sense that the total synergistic DoF-gain can be much larger than the sum of the individual gains from delayed CSIT and from coded caching. Then we incorporate imperfect-quality current CSIT, and explore the interplay between caching and current CSIT quality. In addition to the revealed substantial cache-aided DoF gains, the results suggest the interesting practical ramification that distributing predicted content 'during the night', can offer continuous amelioration of the load of predicting and disseminating CSIT during the day.

In the second part, we explore wireless coded caching from a topological perspective where some users have weaker channels than others. By employing a simple form of interference enhancement, and exploiting the property that weak links attenuate interference, we show that multicasting rates can remain high even when involving weak users. Hence, for coded multicasting, the weak users need not bring down the performance of all users, but on the contrary to a certain extent, the strong users can lift the performance of the weak users without any penalties on their own performance.