A modified-SAW for network selection in heterogeneous wireless networks

Abstract

In the context of heterogeneous networks, users with multi-mode terminals can connect to different radio access technologies such as 802.16, 802.11, HSPA and LTE at the same time. The challenge here is to achieve the Always Best Connected ABC concept; the main issue is the automatic choice of the suitable Radio Access Technology RAT from the list of the available RATs. This decision is called the network selection. In this paper, we propose a modified-SAW function to deal with the drawbacks of the existing solutions. Indeed, the existing MADMs methods suffer mainly from the famous problem of rank reversal once an alternative is added or removed, other problems occur in the legacy MADMs such as the sensitiveness to the user preference in TOPSIS; the penalization of the alternatives with poor attributes values in WPM. Our simulations brought out the lakes of the traditional MADMs approach in the context of network selection. We modify the SAW method intelligently and we use it to solve the network selection problem. Finally, we compare the performance of our solution with the previous works in different scenarios; the simulations show that our proposal outperforms the other existing methods.

A machine learning access network selection in a heterogeneous wireless environment

“…Best-effort services such as email and web browsing do not prioritise any particular quality of service requirements. 15 It is important to explain how the simulations are conducted, and what are we expecting from it. The idea is to compare our proposal (k-means algorithm) with other legacy MADM methods in the case of selecting the best RAT in terms of QoS.…”

“…The weight vector itself is provided in the Table 1. 15 According to Algorithm 1, our proposal necessitates three inputs.…”

“…The range of values for these parameters is displayed in Table 2 and are obtained through simulations. 15,16,22 The simulations were carried out in different scenarios for the different applications (loaded or empty cells, different number of users, users at the middle of the cell and other at the edge where the RSS is very low). This is why we have for each parameter a range of values and for the sake of simulation we take an average value.…”

“…On the other hand, video services such as video calls and streaming prioritise both data rate and low latency, as well as low packet loss, as video comprises audio and images with larger packets. Best‐effort services such as email and web browsing do not prioritise any particular quality of service requirements 15 …”

“…In Reference 15 the authors propose a modified SAW algorithm that deals with the problem of the legacy SAW method. They claim to have the best results compared to the MADM in different scenarios.…”