Exploring Evolutionary Spectral Clustering for Temporal-Smoothed Clustered Cell-Free Networking
Junyuan Wang, Tianyao Wu, Ouyang Zhou, Yaping Zhu
TL;DR
The paper tackles mobility-induced handovers in clustered cell-free networks by jointly maximizing the instantaneous sum rate $R^{(t)}(\\mathcal{M}^{(t)})$ and minimizing handovers via temporal smoothness $S(\\mathcal{M}^{(t)})=R^{(t-1)}(\\mathcal{M}^{(t)})$. It reformulates the problem as a time-varying graph partitioning task on $\\mathcal{G}^{(t)}$ and solves it with evolutionary spectral clustering, using a weighted Laplacian $\alpha \mathbf{L}^{(t)} + (1-\alpha)\mathbf{L}^{(t-1)}$ to balance current and past partitions. The proposed algorithm yields $O(L^3)$ complexity and demonstrates that higher $\alpha$ reduces handovers with small sum-rate degradation compared to a per-time-step maximum baseline. This mobility-aware, scalable clustering approach reduces signaling overhead and improves user experience in dense, dynamic wireless networks.
Abstract
Clustered cell-free networking, which dynamically partitions the whole network into nonoverlapping subnetworks, has been recently proposed to mitigate the cell-edge problem in cellular networks. However, prior works only focused on optimizing clustered cell-free networking in static scenarios with fixed users. This could lead to a large number of handovers in the practical dynamic environment with moving users, seriously hindering the implementation of clustered cell-free networking in practice. This paper considers user mobility and aims to simultaneously maximize the sum rate and minimize the number of handovers. By transforming the multi-objective optimization problem into a time-varying graph partitioning problem and exploring evolutionary spectral clustering, a temporal-smoothed clustered cell-free networking algorithm is proposed, which is shown to be effective in smoothing network partitions over time and reducing handovers while maintaining similar sum rate.