On Cross-Layer Optimization for Real-Time Remote Communication in 5G and Beyond

The use of real-time remote communication has seen significant growth in the last few years. The need for providing the feeling of togetherness in real-time remote communication, combined with new developments in volumetric video, are expected to lead to immersive and holographic remote communication in the near future. These services will require high bandwidth, low latency and significant processing both at the sender and receiver side, and often within the network itself too. In order to fulfill these requirements, we argue that a holistic cross-layer optimization approach, that takes input from and provides optimization actions to all layers involved in the delivery of these services is necessary. In this paper we provide insights in the design and implementation of a cross-layer system orchestrator for adaptation of real-time remote communication. Based on inputs from the network and application layers, it uses a machine learning (ML) model to maximize the objective video quality metric by finding the best system configuration and taking adaptation actions in both layers. The model performance shows that it learned how to offset any system dynamics coming from the environment with the correct configuration settings.