19 February 2025 in IEEE Open Journal of the Communications Society
Cell-free massive MIMO (CF-mMIMO) is a promising approach for future mobile networks, utilizing centralized MIMO processing for densely distributed access points (APs). In CF-mMIMO, to reduce the computational load for signal processing while meeting throughput demands, user equipment (UEs) are served by a selected number of APs. A significant challenge is AP clustering for each UE, particularly in dynamic environments with moving UEs. One approach for optimizing the AP cluster involves deep reinforcement learning (DRL). However, with numerous UEs and APs, the computational load of DRL increases due to the larger model size and higher inference frequency. To address this, we propose an AP clustering method using distributed DRL. The model focuses on determining the AP cluster for every single UE to prevent model size expansion. The per-user models act as distributed actors, enabling parallel inference. Furthermore, to suppress inference frequency, multiple UEs with low mobility are assigned to the same actor, minimizing the number of parallel actors required without compromising throughput. Numerical simulation shows that our proposed method achieves efficient AP clustering that satisfies throughput requirements with reduced computational load in DRL, even in large-scale environments.
1 January 2023 in IEEE Access
20 February 2025 in IEEE Open Journal of the Communications Society
The evolution toward user-centric radio access networks (RANs) necessitates innovative management strategies to handle growing network demands efficiently. This study presents a novel framework for the disaggregated and distributed deployment of near-real-time RAN intelligent controllers (near-RT RICs) tailored for large-scale user-centric networks. By leveraging a mixed-integer nonlinear programming (MINLP) approach, we aim to optimize both control latency and deployment costs, which are critical considerations in next-generation network management. Our proposed divide-and-conquer (D&C) and heuristic-greedy (HG)-based algorithms significantly reduce computational complexity while maintaining near-optimal performance. Simulation results indicated that the distributed framework achieved a 78% reduction in control latency with only 7% increase in deployment cost compared with the traditional centralized approach. Additionally, it achieved 99% reduction in runtime performance with an objective function value that was just 2% lower than the optimal value. These findings confirm the framework’s scalability and effectiveness, validating its suitability for operational telecom environments
4 September 2024 in IEEE Access
In anticipation of the 6G system, it is crucial to address the challenges related to achieving high throughput and low latency over a wide area. This is particularly important for applications such as 4K 360-degree streaming, tactile communication, and autonomous driving. Cell-free massive MIMO (CF-mMIMO) has been proposed as a solution to enhance radio coverage by eliminating cell edges, where a central processing unit (CPU) orchestrates signal processing across distributed access points (APs). This paper explores the deployment of CF-mMIMO within a realistic hierarchical and distributed site architecture to meet the stringent latency requirements of 6G through mobile edge computing (MEC). However, the deployment of CF-mMIMO and utilization of MEC within a distributed and hierarchical architecture necessitates the placement of CPUs at edge sites. This arrangement may lead to radio quality degradation at the edges of distributed sites due to inter-site interference and AP selection limitations. To address this problem, we propose a multi-homing AP connection that allows APs to connect to multiple edge sites, thereby enabling signal processing across sites. Furthermore, we conduct an analysis of the fronthaul (FH) costs associated with multi-homing connections and propose a multi-homing configuration and multi-homing AP selection method that improves radio quality without incurring excessive FH costs. The proposed heuristic algorithm is introduced to optimize the AP connections for multi-homing, giving priority to APs near site edges to strike a balance between cost and performance. Simulation results demonstrate that our proposed method attains near-optimal throughput while preserving cost efficiency.
8 December 2025 in IEEE Access
In beyond fifth-generation (B5G) and sixth-generation (6G) mobile networks, cell-free massive multiple-input multiple-output (CF-mMIMO) has emerged as a promising solution for delivering uniform communication quality across the radio access network. In CF-mMIMO systems, as the number of access points (APs) and user equipment (UEs) increases, the size of the matrices involved in MIMO signal processing grows substantially. These computationally intensive signal processing operations must be completed within the duration of a wireless frame. To alleviate the resulting computational burden, the partial minimum mean squared error (P-MMSE) method processes only selected AP-UE combinations, while hardware accelerators (HAs) such as graphics processing units (GPUs) and field-programmable gate arrays (FPGAs) offer high-speed parallel computation. However, a key inefficiency persists: the fixed matrix dimensions used in parallel operations on HAs led to underutilization of computational resources when processing the smaller matrices generated by P-MMSE. This study introduces an HA-oriented P-MMSE scheme designed to address this limitation by optimizing UE grouping to maximize single-instruction multiple-data (SIMD) set utilization, while preserving the computational efficiency advantages of P-MMSE. Experimental evaluations on a GPU demonstrate that the proposed method improves SIMD set utilization from 62.2% to 75.4% during matrix operations. Under practical conditions with varying AP cluster sizes, it achieves a 53.0% reduction in energy consumption compared to conventional MMSE, thereby validating its effectiveness for energy-efficient wireless signal processing in large-scale CF-mMIMO systems.