ML-based Handover Prediction and AP Selection in Cognitive Wi-Fi Networks

TL;DR

This paper introduces machine learning-based methods for handover prediction and access point selection in cognitive Wi-Fi networks, significantly improving performance over traditional techniques in dense WLAN environments.

Contribution

It presents novel data-driven ML schemes for handover prediction and AP selection that outperform traditional methods in dense Wi-Fi networks.

Findings

Handover prediction reduces unnecessary handovers by 60%.

AP selection achieves throughput gains up to 9.2%.

Proposed schemes outperform traditional signal-based approaches.

Abstract

Device mobility in dense Wi-Fi networks offers several challenges. Two well-known problems related to device mobility are handover prediction and access point selection. Due to the complex nature of the radio environment, analytical models may not characterize the wireless channel, which makes the solution of these problems very difficult. Recently, cognitive network architectures using sophisticated learning techniques are increasingly being applied to such problems. In this paper, we propose data-driven machine learning (ML) schemes to efficiently solve these problems in wireless LAN (WLAN) networks. The proposed schemes are evaluated and results are compared with traditional approaches to the aforementioned problems. The results report significant improvement in network performance by applying the proposed schemes. The proposed scheme for handover prediction outperforms traditional methods i.e. received signal strength method and traveling distance method by reducing the number of unnecessary handovers by 60% and 50% respectively. Similarly, in AP selection, the proposed scheme outperforms the strongest signal first and least loaded first algorithms by achieving higher throughput gains up to 9.2% and 8% respectively.