Use this URL to cite or link to this record in EThOS:
Title: Interference characterisation, location and bandwidth estimation in emerging WiFi networks
Author: Farshad, Arsham
ISNI:       0000 0004 7228 1497
Awarding Body: University of Edinburgh
Current Institution: University of Edinburgh
Date of Award: 2014
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Wireless LAN technology based on the IEEE 802.11 standard, commonly referred to as WiFi, has been hugely successful not only for the last hop access to the Internet in home, office and hotspot scenarios but also for realising wireless backhaul in mesh networks and for point -to -point long- distance wireless communication. This success can be mainly attributed to two reasons: low cost of 802.11 hardware from reaching economies of scale, and operation in the unlicensed bands of wireless spectrum. The popularity of WiFi, in particular for indoor wireless access at homes and offices, has led to significant amount of research effort looking at the performance issues arising from various factors, including interference, CSMA/CA based MAC protocol used by 802.11 devices, the impact of link and physical layer overheads on application performance, and spatio-temporal channel variations. These factors affect the performance of applications and services that run over WiFi networks. In this thesis, we experimentally investigate the effects of some of the above mentioned factors in the context of emerging WiFi network scenarios such as multi- interface indoor mesh networks, 802.11n -based WiFi networks and WiFi networks with virtual access points (VAPs). More specifically, this thesis comprises of four experimental characterisation studies: (i) measure prevalence and severity of co- channel interference in urban WiFi deployments; (ii) characterise interference in multi- interface indoor mesh networks; (iii) study the effect of spatio-temporal channel variations, VAPs and multi -band operation on WiFi fingerprinting based location estimation; and (iv) study the effects of newly introduced features in 802.11n like frame aggregation (FA) on available bandwidth estimation. With growing density of WiFi deployments especially in urban areas, co- channel interference becomes a major factor that adversely affects network performance. To characterise the nature of this phenomena at a city scale, we propose using a new measurement methodology called mobile crowdsensing. The idea is to leverage commodity smartphones and the natural mobility of people to characterise urban WiFi co- channel interference. Specifically, we report measurement results obtained for Edinburgh, a representative European city, on detecting the presence of deployed WiFi APs via the mobile crowdsensing approach. These show that few channels in 2.4GHz are heavily used and there is hardly any activity in the 5GHz band even though relatively it has a greater number of available channels. Spatial analysis of spectrum usage reveals that co- channel interference among nearby APs operating in the same channel can be a serious problem with around 10 APs contending with each other in many locations. We find that the characteristics of WiFi deployments at city -scale are similar to those of WiFi deployments in public spaces of different indoor environments. We validate our approach in comparison with wardriving, and also show that our findings generally match with previous studies based on other measurement approaches. As an application of the mobile crowdsensing based urban WiFi monitoring, we outline a cloud based WiFi router configuration service for better interference management with global awareness in urban areas. For mesh networks, the use of multiple radio interfaces is widely seen as a practical way to achieve high end -to -end network performance and better utilisation of available spectrum. However this gives rise to another type of interference (referred to as coexistence interference) due to co- location of multiple radio interfaces. We show that such interference can be so severe that it prevents concurrent successful operation of collocated interfaces even when they use channels from widely different frequency bands. We propose the use of antenna polarisation to mitigate such interference and experimentally study its benefits in both multi -band and single -band configurations. In particular, we show that using differently polarised antennas on a multi -radio platform can be a helpful counteracting mechanism for alleviating receiver blocking and adjacent channel interference phenomena that underlie multi -radio coexistence interference. We also validate observations about adjacent channel interference from previous studies via direct and microscopic observation of MAC behaviour. Location is an indispensable information for navigation and sensing applications. The rapidly growing adoption of smartphones has resulted in a plethora of mobile applications that rely on position information (e.g., shopping apps that use user position information to recommend products to users and help them to find what they want in the store). WiFi fingerprinting is a popular and well studied approach for indoor location estimation that leverages the existing WiFi infrastructure and works based on the difference in strengths of the received AP signals at different locations. However, understanding the impact of WiFi network deployment aspects such as multi -band APs and VAPs has not received much attention in the literature. We first examine the impact of various aspects underlying a WiFi fingerprinting system. Specifically, we investigate different definitions for fingerprinting and location estimation algorithms across different indoor environments ranging from a multi- storey office building to shopping centres of different sizes. Our results show that the fingerprint definition is as important as the choice of location estimation algorithm and there is no single combination of these two that works across all environments or even all floors of a given environment. We then consider the effect of WiFi frequency bands (e.g., 2.4GHz and 5GHz) and the presence of virtual access points (VAPs) on location accuracy with WiFi fingerprinting. Our results demonstrate that lower co- channel interference in the 5GHz band yields more accurate location estimation. We show that the inclusion of VAPs has a significant impact on the location accuracy of WiFi fingerprinting systems; we analyse the potential reasons to explain the findings. End -to -end available bandwidth estimation (ABE) has a wide range of uses, from adaptive application content delivery, transport-level transmission rate adaptation and admission control to traffic engineering and peer node selection in peer -to- peer /overlay networks [ 1, 2]. Given its importance, it has been received much research attention in both wired data networks and legacy WiFi networks (based on 802.11 a/b /g standards), resulting in different ABE techniques and tools proposed to optimise different criteria and suit different scenarios. However, effects of new MAC/PHY layer enhancements in new and next generation WiFi networks (based on 802.11n and 802.11ac standards) have not been studied yet. We experimentally find that among different new features like frame aggregation, channel bonding and MIMO modes (spacial division multiplexing), frame aggregation has the most harmful effect as it has direct effect on ABE by distorting the measurement probing traffic pattern commonly used to estimate available bandwidth. Frame aggregation is also specified in both 802.11n and 802.1 lac standards as a mandatory feature to be supported. We study the effect of enabling frame aggregation, for the first time, on the performance of the ABE using an indoor 802.11n wireless testbed. The analysis of results obtained using three tools - representing two main Probe Rate Model (PRM) and Probe Gap Model (PGM) based approaches for ABE - led us to come up with the two key principles of jumbo probes and having longer measurement probe train sizes to counter the effects of aggregating frames on the performance of ABE tools. Then, we develop a new tool, WBest+ that is aware of the underlying frame aggregation by incorporating these principles. The experimental evaluation of WBest+ shows more accurate ABE in the presence of frame aggregation. Overall, the contributions of this thesis fall in three categories - experimental characterisation, measurement techniques and mitigation/solution approaches for performance problems in emerging WiFi network scenarios. The influence of various factors mentioned above are all studied via experimental evaluation in a testbed or real - world setting. Specifically, co- existence interference characterisation and evaluation of available bandwidth techniques are done using indoor testbeds, whereas characterisation of urban WiFi networks and WiFi fingerprinting based location estimation are carried out in real environments. New measurement approaches are also introduced to aid better experimental evaluation or proposed as new measurement tools. These include mobile crowdsensing based WiFi monitoring; MAC/PHY layer monitoring of co- existence interference; and WBest+ tool for available bandwidth estimation. Finally, new mitigation approaches are proposed to address challenges and problems identified throughout the characterisation studies. These include: a proposal for crowd - based interference management in large scale uncoordinated WiFi networks; exploiting antenna polarisation diversity to remedy the effects of co- existence interference in multi -interface platforms; taking advantage of VAPs and multi -band operation for better location estimation; and introducing the jumbo frame concept and longer probe train sizes to improve performance of ABE tools in next generation WiFi networks.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available