Quality of experience measurement, prediction and provisioning in heterogeneous access networks

Abstract: In mobile computing systems, users can access network services anywhere and anytime using mobile devices such as tablets and smart phones. Users usually have some expectations about the services provided to them by different service providers, for example, telecommunication operators and network providers. Users’ expectations along with additional factors such as cognitive and behavioural states, cost, and network quality of service (QoS) may determine their quality of experience (QoE). If users are not satisfied with their QoE, they may switch to different providers or may stop using a particular application or service. QoE measurement and prediction can benefit users to avail personalized services from service providers. On the other hand, it can help service providers to achieve lower user-operator switch-over. Users with mobile devices can roam in heterogeneous access networks (HANs). A mobile device may go through handoffs while roaming in HANs i.e., it may switch from one access point (AP) to another. These APs within a network can belong to different network technologies, for example, WLAN or 4G. Handoffs may cause severe QoE degradation due to increased delay and packet losses. Thus, there is a need to facilitate QoE-aware handoffs for users roaming in HANs. The mobile devices can learn from the prior network conditions and users' QoE to make timely and proactive QoE-aware handoffs.In this thesis, we propose, develop and validate a novel method, CaQoEM-Context-aware Quality of Experience, Modelling, Measurement and Prediction. CaQoEM is based on Bayesian networks and utility theory. It provides a straightforward and efficient way of dealing with a plethora of parameters to model, measure and predict QoE under uncertainty on a single scale. We validate CaQoEM using a number of case studies, user tests and simulations performed in OPNET. Our results validate that CaQoEM can efficiently model, measure and predict users' QoE. It achieves an average QoE prediction accuracy of 98.93% in stochastic wireless network conditions such as wireless signal fading, handoffs and wireless network congestion. We further extend our CaQoEM to develop SCaQoEM-Sequential Context-aware Quality of Experience Measurement and Prediction where we use dynamic Bayesian networks and utility theory to model, measure and predict users’ QoE over time. We performed a case study and our results validate the efficiency of SCaQoEM.In this thesis, we also propose, develop and validate a novel approach called PRONET-Proactive Context-aware Mobility Support in HANs. PRONET incorporates a novel method for QoE estimation and prediction using hidden Markov models and Multi-homed Mobile IP. It also incorporates a method for QoE-aware handoffs using Q-learning function. Using extensive simulations and experimental analysis, we show that PRONET achieves an average QoE prediction accuracy of 97%. Further, PRONET can maximize users’ QoE by reducing the average number of handoffs by 60.65%, compared to the state-of-the-art methods. The outcomes of this thesis have resulted in eleven peer-reviewed conference, workshop and journal papers along with three technical reports.