Virtual assistant mediated communications for radio resources saving

Artificial Intelligence (AI) and Machine Learning (ML) are considered by 3GPP and ITU important technologies to manage and control the radio access networks (RANs). The efficient usage of radio resources is one of the objectives of the network AI/ML-based management/control system. The possibility of passing information on the generated traffic to the network can be helpful in optimizing the usage of radio resources allocated to users to connect with the application servers in the cloud and/or in the edge cloud. In this paper, we discuss one approach to save radio resources. To this purpose, we consider the possibility of splitting the user-to-application server link into two cascaded links inter-connected by an AI-based virtual assistant (VA). The first link, from the user to the VA, is within the RAN while the second one, from the VA to the application server is inside the Telco network. Assuming the VA could impersonate the user by exchanging messages with the application servers, at the end of transactions only the net data fulfilling the user request are transferred to the user over the RAN. We refer to this approach as VA-mediated communications. A description of the network architecture including VAs is presented in this paper. In general, depending on the type of the invoked application/service, this approach permits to reduce the amount of traffic transmitted over the RAN. We analyze the performance improvement in terms of radio resource saving by considering two applications: the download of a web page and the access to a web application. It is observed that the achievable radio resource saving depends on the ability of the VA to impersonate the user behavior while interacting with the application server(s). We provide a model for the evaluation of the additional costs of the hardware/software for the Telco to integrate the VA infrastructure into its network i.e. in the cloud or in the edge cloud.