User-Access Point Association for High Density MIMO Wireless LANs

TL;DR

This work tackles interference and contention in dense UL-MIMO WLANs by formulating user-AP association as a throughput-utility optimization that integrates CSI, MAC-layer effects, and interference. It recasts the problem as a maximum weighted bipartite graph matching and solves it with the Kuhn-Munkres algorithm, complemented by a dynamic variant for new entrants and mobility. Simulations show average aggregate-throughput gains of about $36.9 ext{ ext%}$ over SSF, $33.5 ext{ ext%}$ over Greedy, $20.4 ext{ ext%}$ over SmartAssoc, and $11.3 ext{ ext%}$ over BPF, with larger improvements in dense networks and notable per-user improvements. The dynamic graph-based approach (GDA) offers robust adaptation with reduced complexity in dynamic settings, making the method practically attractive for deploying high-density UL-MIMO WLANs.

Abstract

Wireless local area network (WLAN) access points (APs) are being deployed in high density to improve coverage and throughput. The emerging multiple-input multiple-output (MIMO) implementation for uplink (UL) transmissions promises high per-user throughput and improved aggregate network throughput. However, the high throughput potential of dense UL-MIMO WLAN is impaired by multiple access channel interference and high contention among densely distributed user stations (STAs). We investigate the problem of actualizing the throughput potential of UL-MIMO in high density WLANs via user-AP association. Since user-AP association influences interference and STA contention, a method to optimally distribute STAs among APs is proposed to maximize aggregate users' throughput utility. This problem is transformed into a graph matching problem with the throughput utility function as the graph edge weights. The graph matching problem is solved as a combinatorial problem using a modified classical Kuhn-Munkres algorithm. A dynamic implementation of the proposed algorithm is used to periodically update user-AP associations when there are changes in the network due to new entrants and/or user mobility. Simulated dense UL-MIMO WLAN scenarios reveal that the proposed scheme achieves an average of $36.9 \%$, $33.5 \%$, $20.4 \%$ and $11.3 \%$ gains over the default strongest signal first (SSF) association scheme used in conventional WLAN, Greedy [14], SmartAssoc [13] and best performance first (BPF) [5] algorithms, respectively.

Figures (10)

• Figure 1: Uplink MU-MIMO model: $U$ antennas STA and $K$ antennas AP.
• Figure 2: Channel negotiation procedure.
• Figure 3: Graph representation of a wireless LAN.
• Figure 4: Sample realization of a network with an AP density $\eta_m = 0.2$ and an STA density $\eta_n = 0.8$: $M = |\mathcal{A}|= 35$, $N = |\mathcal{N}|= 169$.
• Figure 5: Aggregate throughput versus network size with $K = 8$ antennas at the APs, $U = 4$ transmit antennas at the STAs, fairness index $\delta = 0.5$ and $M = |\mathcal{A}| = 35$.

Theorems & Definitions (3)

• Remark 1
• Definition 1
• Remark 2