Multiobjective planning of indoor Wireless Local Area Networks using subpermutation-based hybrid algorithms

Abstract

Wireless Local Area Network (WLAN) has become the most popular technology for mobile Internet access in recent decades. This manuscript presents a novel approach, based on hybrid optimization algorithms, for planning WLANs. Two objective functions are optimized: to maximize network load balance and signal-to-noise ratio. In addition, constraints related to coverage, customer, and equipment demand are considered. A key aspect of the proposed algorithm is its new representation/decoding scheme, based on subpermutations, which considerably reduces the search space dimension. This structure guarantees feasibility of the obtained solutions and increases the computational efficiency of the method. Several tests were performed in two scenarios, one of them using real data from a large-scale WLAN. When compared to other three approaches, such results show that the proposed method provides solutions that reduce costs and improve the WLAN throughput.

Introduction

Over the past decade, people have become more dependent on wireless communications. The widespread adoption of Wireless Local Area Networks (WLANs) has been motivated by mobility, high data throughput, and low structural cost. Thus, the potential arising from the massive presence of access points (APs) has boosted the application of this technology, which goes beyond providing Internet connection for accessing Web pages, video streaming services, and social networks. In the 5G era, the IEEE 802.11 standard is becoming another access technology option to data offload, relieving the overhead in cellular networks. Such characteristics highlight the relevance of WLAN planning facing emerging demands.

WLAN planning is often driven to provide suitable signal coverage in some targeted areas. However, key aspects such as interference level, AP capacity, and distribution of customers in the covered area are frequently neglected, despite their importance. These factors become critical in applications that require stable and low-delay connections, such as videoconferencing, voice over IP, and multiplatform instant messaging. Nowadays, these requirements have become essential with the BYOD (Bring Your Own Device) trend, in which users demand high-performance wireless networks on their mobile devices practically all the time.

Overprovisioning APs, besides requiring more investment, also can lead to degradation of network performance, due to interference problems caused by the limited frequency spectrum available in WLANs [1]. Therefore, APs must be located to meet network coverage and client demands, while minimizing interference between the access points. In real situations, many users tend to agglomerate in the same area [2], [3]. Such a behavior, combined with the increase in demand for different services, creates an unbalanced load on the wireless LAN, which may compromise its overall performance and the quality of service. Again, a suitable AP placement, combined with an efficient client-AP association policy, is important to maximize the signal-to-noise ratio on the client devices and improve the load balance.

The aforementioned aspects make WLAN planning a hard optimization problem, in which multiple design criteria and technical constraints must be considered. The application of exact methods to solve this problem is not feasible, given their high computational cost [4], [5]. Thus, metaheuristics appear as alternatives to obtain reasonable solutions within acceptable computational time. Genetic Algorithms (GAs), for instance, have been used to solve WLAN related problems [6], [7], [8], [9], [10]. Furthermore, metaheuristics are scalable and can be adapted to deal with problems of different features and dimensions.

An innovative approach, based on hybrid algorithms, is proposed in this work for Wi-Fi network planning. Such an approach is scalable and combines two evolutionary algorithms and one greedy method to obtain efficient designs for large-scale WLANs, considering multi-floor buildings. The developed method uses a new efficient mechanism for representing and decoding solutions by breaking the original permutation into subpermutations. This structure guarantees solution feasibility and promotes a dimensional reduction by preventing the algorithm from performing unnecessary operations. Two design criteria are considered: maximizing load balance and signal-to-interference-and-noise (SINR) ratio. The generated solutions should satisfy constraints of minimum coverage, availability of channels, and bandwidth of the APs. Two test scenarios were considered, with one of them built with real data collected from a large ESS (Extended Service Set) WLAN already in operation. Finally, the solutions were submitted to a sensitivity analysis procedure to validate their robustness regarding user profile variations (locations and demands) and access point failures.

This paper is organized as follows. Section 2 discusses some related works. Section 3 overviews the problem of WLAN planning, and Section 4 addresses its mathematical model. Section 5 presents the proposed approach, Section 6 states test scenarios, and Section 7 evaluates the performance of the proposed method on those scenarios. Finally, Section 8 concludes the manuscript.