Multiple-antenna Placement Delivery Array for Cache-aided MISO Systems
Abstract
We consider the cache-aided multiple-input single-output (MISO) broadcast channel, which consists of a server with antennas and single-antenna users, where the server contains files of equal length and each user is equipped with a local cache of size files. Each user requests an arbitrary file from library. The objective is to design a coded caching scheme based on uncoded placement and one-shot linear delivery, to achieve the maximum sum Degree-of-Freedom (sum-DoF) with low subpacketization. It was shown in the literature that under the constraint of uncoded placement and one-shot linear delivery, the optimal sum-DoF is . However, previously proposed schemes for this setting incurred either an exponential subpacketization order in , or required specific conditions in the system parameters , , and . In this paper, we propose a new combinatorial structure called multiple-antenna placement delivery array (MAPDA). Based on MAPDA and Latin square, the first proposed scheme achieves the optimal sum-DoF with the subpacketization of when . Subsequently, for the general case we propose a transformation approach to construct an MAPDA from any -regular PDA (a class of PDA where each integer in the array occurs times) for the original shared-link coded caching problem. When the original PDA corresponds to the Maddah-Ali and Niesen coded caching scheme, the resulting scheme under the combinatorial structure of MAPDA can achieve the optimal sum-DoF with reduced subpacketization with respect to the existing schemes. The work can be extended to the multiple independent single-antenna transmitters (servers) corresponding to the cache-aided interference channel proposed by Naderializadeh et al. and the scenario of transmitters equipped with multiple antennas.
Cite
@article{arxiv.2201.11462,
title = {Multiple-antenna Placement Delivery Array for Cache-aided MISO Systems},
author = {Ting Yang and Kai Wan and Minquan Cheng and Giuseppe Caire},
journal= {arXiv preprint arXiv:2201.11462},
year = {2022}
}
Comments
33 pages