Reduced complexity multicast beamforming and group assignment schemes for multi-antenna coded caching

Abstract. In spite of recent advancements in wireless communication technologies and data delivery networks, it is unlikely that the speeds supported by these networks will be able to keep up with the exponentially increasing demand caused by the widespread adoption of high-speed and large-data applications. One appealing idea proposed to address this issue is coded caching, which is an innovative data delivery technique that makes use of the network’s aggregate cache rather than the individual memory available to each user. This proposed idea of coded caching helps boost the data rates by distributing cache material throughout the network and delivering independent content to many users at a time. Despite the original theoretical promises for large caching gains, in reality, coded caching suffers from severe bottlenecks that dramatically limit these gains. Some of these bottlenecks are requiring complex successive interference cancellation (SIC) at the receiver, exponential increase in subpacketization, applicability to a limited range of input parameters, and performance losses in low- and mid- signal to noise ratio (SNR) regimes. In this study, we present a novel coded caching scheme based on user grouping for cache-aided multi-input single-output (MISO) networks. One special property of this new scheme is its applicability to every set of input values for the user count ($K$), transmitter-side antenna count ($L$), and the global coded caching gain ($t$). Moreover, for a fixed $t$, this scheme can achieve theoretical sum-DoF optimality with no limitations. This strategy yields superior performance in terms of subpacketization when input parameters satisfy $\frac{t+L}{t+1} \in \mathbb{N}$. This performance boost is enabled by the underlying user grouping structure during data delivery. However, when input parameters do not comply with $\frac{t+L}{t+1}$ $\in \mathbb{N}$, in order to guarantee symmetry of the scheme and optimal DoF, multicast and unicast messages need to be constructed using a tree diagram, resulting in excess subpacketization and transmission count. Nevertheless, the simple receiver structure without the SIC requirement not only simplifies the implementation complexity but also enables us to use state-of-the-art methods to readily design optimized transmit beamformers maximizing the achievable symmetric rate. Finally, we use numerical analysis to compare our new proposed scheme with well-known coded caching schemes in the literature