Antti Tölli

Takumi Takahashi, Antti Tölli, Shinsuke Ibi et al.

This paper proposes a novel layered belief propagation (BP) detector with a concatenated structure of two different BP layers for low-complexity large multi-user multi-input multi-output (MU-MIMO) detection based on statistical beams. To reduce the computational burden and the circuit scale on the base station (BS) side, the two-stage signal processing consisting of slow varying outer beamformer (OBF) and group-specific MU detection (MUD) for fast channel variations is effective. However, the dimensionality reduction of the equivalent channel based on the OBF results in significant performance degradation in subsequent spatial filtering detection. To compensate for the drawback, the proposed layered BP detector, which is designed for improving the detection capability by suppressing the intra- and inter-group interference in stages, is introduced as the post-stage processing of the OBF. Finally, we demonstrate the validity of our proposed method in terms of the bit error rate (BER) performance and the computational complexity.

Mohammad NaseriTehrani, MohammadJavad Salehi, Antti Tölli

This is an extended journal version of the conference paper published in ISIT 2025; submitted to IEEE Transactions on Communications (TCOM). Integrating coded caching (CC) into multiple-input multiple-output (MIMO) communications significantly enhances the achievable degrees of freedom (DoF). This paper investigates a practical cache-aided asymmetric MIMO configuration with cache ratio $γ$, where a server with $L$ transmit antennas communicates with $K$ users. The users are partitioned into $J$ groups, and each user in group $j$ has $G_j$ receive antennas. We propose four content-aware MIMO-CC strategies: \emph{min-$G$} enforces symmetry using the smallest antenna count among users; \emph{Grouping} maximizes intra-subset spatial multiplexing gain at the expense of some global caching gain; \emph{Super-grouping} aggregates users into optimized \emph{min-$G$}-based super-sets with identical effective receive multiplexing gains before applying \emph{Grouping} across them; and \emph{Phantom} redistributes spatial resources assuming ``phantom'' antennas at the users to bridge the performance gains of \emph{min-$G$} and \emph{Grouping}. We develop these asymmetric strategies under three reference symmetric CC placement-delivery policies with guaranteed linear decodability: a DoF-optimal policy achieving the optimal single-shot DoF, and two closed-form policies, namely combinatorial and linear cyclic low-complexity constructions, with the cyclic policy attaining DoF performance close to the others in many operating regimes. Analytical and numerical results demonstrate significant DoF improvements across various system configurations, and that policy-strategy combinations offer flexible trade-offs between DoF and subpacketization complexity.