Resource Allocation for Multi-Antenna Coded Caching Systems with Dynamic User Behavior

In practical cache-aided setups, users may enter or depart the network at any time. The shared caching model can mitigate the detrimental impact of such dynamic user behavior by assigning users to a limited set of caching profiles. In this paper, we propose a highly effective data delivery strategy relying on an uneven subpacketization process to improve the finite-SNR performance of dynamic networks. In order to maximize the utilization of available spatial degrees of freedom, the number of serving users per transmission is fixed to the maximal value, resulting in a larger number of transmissions for users assigned to caching profiles with smaller user counts. Consequently, users assigned to different profiles may receive a varying portion of their requested files in each transmission. To facilitate the efficient delivery of unequal-sized data, an iterative beamformer design given a variable rate allocation per user is also devised. Numerical results demonstrate that the performance of the proposed method is always superior to unicasting and other existing coded caching approaches.