Code, Cache and Deliver on the Move: A Novel Caching Paradigm in Hyper-Dense Small-Cell Networks

TLDR: This work revisits the caching problem in realistic environments where moving users intermittently connect to multiple SBSs encountered at different times and introduces an optimization framework that models user movements via random walks on a Markov chain aimed at minimizing the load of the macro-cell.

Abstract: Caching popular content files at small-cell base stations (SBSs) has emerged as a promising technique to meet the overwhelming growth in mobile data demand. Despite the plethora of work in this field, a specific aspect has been overlooked. It is assumed that all users remain stationary during data transfer and therefore a complete copy of the requested file can always be downloaded by the associated SBSs. In this work, we revisit the caching problem in realistic environments where moving users intermittently connect to multiple SBSs encountered at different times. Due to connection duration limits, users may download only parts of the requested files. Requests for files that failed to be delivered on time by the SBSs are redirected to the coexisting macro-cell. We introduce an optimization framework that models user movements via random walks on a Markov chain aimed at minimizing the load of the macro-cell. As the main contribution, we put forward a distributed caching paradigm that leverages user mobility predictions and innovative information-mixing methods based on the principle of network coding. Systematic experiments based on measured traces of human mobility patterns demonstrate that our approach can offload $65 \text{percent}$ more macro-cell traffic than existing caching schemes in realistic settings.