Autonomic reconfigurable 5G network slicing enabling immersive VR applications in 5G&B softwarized networks

In recent years, telecommunications networks have undergone a marked transformation, both in terms of role and architecture. The need to meet the increasingly stringent requirements of new applications, such as extended reality, has necessitated the introduction of the network programmability. It is not possible to design one network for all applications; instead, we must ensure that each application perceives the network as uniquely dedicated to it and capable of meeting its performance requirements. Programmability, through concepts such as network slicing introduced with 5G, SDN, NFV, and AI, plays a fundamental role in ensuring that the network is easily reconfigurable. Alongside traditional optimization models and the use of AI for network design, there is also the need for a preliminary study of the applications that will use a particular network slice, to obtain a preliminary estimate of the final performance. To this end, in this paper we leverage network softwarization provided by new 5G&B networks, and introduce an Adaptive Closed-loop Encoding VF named SVEs for adaptive compression of 360∘ video streaming. This VF is able to apply a hierarchical compression taking into account both the bandwidth available in the network, and the user viewport. The agent in charge of deciding the different compression ratios uses Deep Reinforcement Learning to optimize a reward function and adapt to the changes in the end-to-end latency, the user movements, and the video content. The numerical results represent an excellent starting point to allow an autonomic reconfiguration of network slice.