Distributed Hybrid Precoding for Edge-Computing-Assisted Cell-Free Massive MIMO Systems With Local CSI

Precoding technology is very promising in edge-computing-assisted cell-free massive multiple-input-multiple-output (ECF-mMIMO) systems because it can eliminate interference and thus improve performance. However, it is challenging to design a computationally efficient hybrid precoding scheme to maximize the achievable rate when only the local channel state information (CSI) is known. To maximize the achievable rate and minimize the computational energy consumption within the tolerable computational latency, we propose a novel optimization framework for joint design of distributed hybrid precoding and computational offloading decision in ECF-mMIMO systems. The joint optimization problem is modeled as maximizing the sum of the ratio of the achievable rate to the computational energy consumption. Since it is difficult to directly solve the joint optimization problem with nonconvex and fractional constraints, we first use the quadratic transform method to remove the fractional constraint and perform an equivalent transformation of the original problem. Next, we decompose the equivalent optimization problem into three subproblems and propose an alternate optimization algorithm. Specifically, we successively adopt a local block diagonalization hybrid precoding scheme and a game-based power allocation algorithm to maximize the total rate, and a computational resource allocation scheme to minimize computational offloading energy consumption under the constraint of computational latency. Numerical simulation results illustrate that the performance obtained by our proposed scheme with only local CSI is 93% of that as compared to the full CSI case. Besides, we also prove the convergence of the alternate optimization algorithm theoretically.