Design and implementation of AR/VR services at the edge infrastructure using Network Function Virtualization

Nowadays Augmented Reality (AR) and Virtual Reality (VR) are dominant IT research and industry related topics. They exploit on-device advanced technology, based on Machine Learning (ML) algorithms, to add useful information above the real world (AR) or to create a complete new virtualized world where the user takes an important role (VR). However, AR and VR, although they are equipped with common low-cost peripherals (IP cameras, LCD displays, speakers, microphones, etc.), they need special purpose core components to run ML power-demanding algorithms which dramatically increase the overall cost of the device. Moreover, those devices lack of portability and flexibility, since they are usually designed for a specific service, thus making impossible the re-use of the device for another task. The goal of this thesis is to present a solution that allows to provide AR/VR services requested by a client using legacy devices, without affecting implementation of the latter. We introduce Network Edge AR/VR (NEAR), a NFV framework for AR/VR services deployed transparently at the edge. The edge infrastructure, namely a network area provided with high-computing power and located near the customer requesting a service, is needed to meet the ultra-low latency requested by those services. Furthermore, a NFV framework is useful wherever we want to optionally provide flexibility and scalability to the requested services, such as content caching and client APIs to dynamically manage the running service. NEAR's philosophy is to offload computing capabilities from a special-purpose device (Server) to a general-purpose Edge Gateway acting as a Proxy server allowing a client to request the augmented/virtual service trough the proxy, in charge of performing the selected service. Transparency, a key feature of NEAR, avoids to affect design and implementation of the endpoints (server and client). In fact, NEAR does not need special equipment and/or specific components both at client-side and server-side in order to perform the chosen AR/VR task.