Application-Specific Routing for Managed Wireless Access

In recent years, devices with wireless capabilities are in particular interest, while the number of mobile devices has been predicted to grow significantly in upcoming years. Besides, emerging technologies particularly the smart home/smart building and e-health have posed new requirements in terms of reliability, predictability, and quality of service in wireless infrastructure. Thus, without a certain level of quality of service, the usability of new technologies cannot be achieved and at the same time, it results in a negative user experience. In order to enhance the wireless services, the quality needs to be predictable and measurable. This project aimed to identify methods and practices for measuring and predicting the wireless quality of experience. This project studied the user expectations and requirements to correlate them with technical parameters in wireless infrastructure so that it established a baseline for predictable and measurable quality of experience. On this basis, this project proposes a novel perceived quality of service (PQoS) assessment using machine learning (ML) methodology called PQoSML to classify the perceived quality of internet services and Wi-Fi networks in homes and buildings. In the proposed methodology, ML techniques produce an interpretable ML model that accurately correlates network performance parameters to the perceived quality of individual internet services or the entire Wi-Fi network. Furthermore, the ML model is computationally efficient, which is executable in the form of edge computing on off-the-shelf Wi-Fi access points (APs) on customer premises resulting in reducing the cost of data analytics while preserving customers' privacy. In effect, the results of this project enable internet service providers (ISPs) to replace traditional QoS monitoring with an efficient wireless quality assessment (QA) strategy capable of predicting user experience in Wi-Fi networks. Therefore, the results of this project provide an applicable solution for ISPs to deliver predictable and measurable wireless quality in homes and buildings.

Piloting wireless services in this project demonstrated the accuracy, interpretability and computational efficiency of our novel PQoSML methodology for predicting the perceived quality of Wi-Fi networks. This methodology empowers human operators to interpret the decisions path in the ML model and thereby infer the probable root causes of quality degradation in Wi-Fi networks. Executing the machine learning (ML) model on off-the-shelf Wi-Fi APs reduces the operational costs of data analytics for internet service providers (ISPs) while protecting customers' privacy by limiting the amount of data sent to the cloud infrastructure. In effect, the quality assessment (QA) strategy proposed in this project enables ISPs to replace traditional quality assessments approaches with an efficient wireless QA strategy capable of predicting user experience in Wi-Fi networks. Thus, it is an applicable solution for ISPs to deliver predictable and measurable wireless quality in homes and buildings.

The predicted explosion in machine-to-machine communication, "smart homes", "e-health", and the Internet of Things, all rely on a crucial, yet elusive element: Predictable wireless service quality. While current infrastructures are limited in terms of (remote) configurability and service differentiation, smart infrastructures will use a service-oriented view. This is a novel view where services, devices and systems will interconnect to provide added value or functionality, for example in the case of trusted connections for medical data, or secure and reliable communications in the case of an alarm. One of the major challenges in smart infrastructures such as home and buildings is the integration of applications that have a wide range of requirements. Current wireless solutions are based on best-effort Wi-Fi networks, and control of the network quality ends at the termination in the home. For the digitised life there is a requirement for standardised methods of seamless exchange of information and data between different types of infrastructure, systems and applications. The project will use different types of technologies and tools in order to automate such as machine learning, analytics, measurements with a combination of scientific methods to find proof of concept of how to predict quality in a wireless network and environment. Eye Networks currently delivers wireless mesh networks for the home (and small businesses) in the form of wireless access points from vendor AirTies, and an ISP service for managing these mesh networks called EyeSaaS Pearl, on our own service platform EyeSaaS. Eye Networks collaborates closely with AirTies as well as with our ISP customers, providing this project with a unique starting point and access to representatives for large parts of the affected value chain.