Clustered cell-free networking is a promising paradigm for future mobile communication systems, which dynamically partitions the whole network into multiple small subnetworks to avoid the cell-edge problem in cellular networks. To optimize network partition, previous approaches primarily relied on heuristics and relaxation techniques. Recent studies leveraged graph partitioning theory to address this problem by representing the wireless network as an undirected bipartite graph. In this paper, we focus on the load balanced clustered cell-free networking problem with the objective of maximizing the minimum sum rate among all subnetworks. In contrast to previous works, we propose a new directed graph model and equivalently transform the problem into a sequence of min-max K-cut problems. Subsequently, a streaming balanced assignment algorithm is proposed to solve min-max K-cut problems. Building upon this, we develop a sequential min-max K-cut approach with theoretical guarantees. Simulation results demonstrate that our method outperforms existing algorithms by significantly improving the minimum subnetwork sum rate, thereby effectively balancing the loads of subnetworks.

The fast development of Artificial Intelligence (AI) agents provides a promising way for the realization of intelligent and customized wireless networks. In this paper, we propose a Wireless Multi-Agent System (WMAS), which can provide intelligent and customized services for different user equipment (UEs). Note that orchestrating multiple agents carries the risk of malfunction, and multi-agent conversations may fall into infinite loops. It is thus crucial to design a conversation topology for WMAS that enables agents to complete UE task requests with high accuracy and low conversation overhead. To address this issue, we model the multi-agent conversation topology as a directed acyclic graph and propose a reinforcement learning- based algorithm to optimize the adjacency matrix of this graph. As such, WMAS is capable of generating and self-optimizing multi-agent conversation topologies, enabling agents to effectively and collaboratively handle a variety of task requests from UEs. Simulation results across various task types demonstrate that WMAS can achieve higher task performance and lower conversation overhead compared to existing multi-agent systems. These results validate the potential of WMAS to enhance the intelligence of future wireless networks.

Clustered cell-free networking that dynamically decomposes a network into multiple subnetworks is emerging as a promising solution to the cell-edge problem in mobile communication systems. Maximizing the minimum subnetwork sum ergodic capacity is of great importance for balancing the loads of different subnetworks. Such a load-balanced clustered cell-free networking problem was proposed recently, yet still lacks efficient algorithms due to its complicated objective function and combinatorial nature. Recognizing these challenges, this paper proposes a tight relaxation method to equivalently transform the original problem into a continuous one that shares the same optimal solution. The relaxed problem is a nonconvex-linear minmax problem, which can be solved by finding its Nash equilibrium. We then propose an alternating gradient projection (AGP) algorithm to effectively tackle it, with a guaranteed convergence. Simulation results show that our approach significantly outperforms the existing benchmarks.

This paper presents a novel sequential max k-cut approach to the pilot assignment problem in cell-free networks. In contrast to the existing max k-cut formulations that were proposed based on intuitions or qualitative analyses, we first establish a theoretical connection between pilot assignment and max k-cut, which avoids the performance degradation brought by the inaccurate max k-cut formulation. Specifically, we employ optimization techniques to equivalently transform the pilot assignment problem for uplink throughput maximization into a series of max k-cut problems with updated weights, leading to a sequential max k-cut approach. Moreover, different from the existing works that decouple pilot assignment and power control, this approach enables us to jointly optimize pilot assignment and power control with a customized scheme developed for further improvement of overall throughput and user fairness. Simulation results show the effectiveness and efficiency of the proposed approach, exhibiting significant improvement over existing methods. This sequential max k-cut approach could serve as a promising candidate for mitigating the performance deterioration resulting from severe pilot reuse in future ultra-dense cell-free networks.

Determining the network capacity, which is a crucial performance metric of wireless systems, is becoming increasingly important with the growing need for future ultra-dense networks. There have been a multitude of works applying random matrix theory (RMT) to capacity analysis. However, most of them approximate the interference as noise and rely on the selection of hyper-parameters, and thus impairs the accuracy. In this paper, we first propose a double splitting model to decompose the capacity into four parts, two of which can be analytically calculated, while the other two are significantly smaller and thus have minimal impact on the overall accuracy. This helps to avoid the approximations of previous methods, simplifying the calculation of capacity and improving the numerical stability. Second, to compute the aforementioned smaller parts, we generalize the moment passing method to more scenarios, and avoid the hyper-parameter selection that impairs the robustness. We also derive the recursive expressions of the moments of any order, enabling flexible trade-offs between efficiency and accuracy. Numerical experiments demonstrate the high efficiency and accuracy of our methods.

Clustered cell-free networking is a promising networking scheme for future mobile communications systems where the base-stations (BSs) are densely deployed. Despite its great importance, finding the optimal networking strategy aiming at maximizing the sum rate of users in the network is a non-convex combinatorial optimization problem. Previous work relaxed the clustered cell-free networking problem into a graph min$K$-cut problem to solve it suboptimally. In this paper, we leverage optimization techniques to equivalently transform the original problem into a series of graph min$K$-cut problems with theoretical guarantee. It is worth mentioning that our approach is highly general, as it is applicable to various constraints, offering adaptability and flexibility to diverse practical networking scenarios. We apply this approach to three typical clustered cellfree networking problems. Simulation results show a consistent improvement of our approach compared to existing algorithms.

The integration of artificial intelligence (AI) and mobile networks is regarded as one of the most important scenarios for 6G. In 6G, a major objective is to realize the efficient transmission of task-relevant data. Then a key problem arises, how to design collaborative AI models for the device side and the network side, so that the transmitted data between the device and the network is efficient enough, which means the transmission overhead is low but the AI task result is accurate. In this paper, we propose the multi-link information bottleneck (ML-IB) scheme for such collaborative models design. We formulate our problem based on a novel performance metric, which can evaluate both task accuracy and transmission overhead. Then we introduce a quantizer that is adjustable in the quantization bit depth, amplitudes, and breakpoints. Given the infeasibility of calculating our proposed metric on high-dimensional data, we establish a variational upper bound for this metric. However, due to the incorporation of quantization, the closed form of the variational upper bound remains uncomputable. Hence, we employ the Log-Sum Inequality to derive an approximation and provide a theoretical guarantee. Based on this, we devise the quantized multi-link information bottleneck (QML-IB) algorithm for collaborative AI models generation. Finally, numerical experiments demonstrate the superior performance of our QML-IB algorithm compared to the state-of-the-art algorithm.

Clustered cell-free networking emerges as a promising paradigm for future communication systems. To optimize clustered cell-free networking, we aim to maximize the sum ergodic capacity, which stands as a key performance metric of wireless communications systems. However, the networking aspect leads to a difficult combinatorial problem. Moreover, the implicitness of ergodic capacity results in obstacles to evaluation and optimization. Existing works often rely on inaccurate approximations of ergodic capacity and/or high-cost integer programming (IP) algorithms, rendering them impractical for real-world deployment. In this paper, we formulate the problem as an IP problem and relax it into a continuous one with the ergodic capacity approximated by its deterministic equivalents. Importantly, we establish the tightness of the relaxation without sacrificing the optimality of the solution. We then employ Breg-man proximal gradient (BPG) nested with Dykstra's algorithm to solve the relaxed problem and show the convergence for both BPG and its subproblems. Simulation results verify the effectiveness and efficiency of our approach.