Adaptive Server Control for Low-Latency Applications over the Cellular Network

Sammanfattning: Queue latency is a fundamental part of end-to-end latency and ensuring low end-to-end latency is crucial for a good quality of service in the upcoming 5G and 6G applications. In this work we present a model capturing the essential dynamics of an end-to-end cellular system upon which we derive a control structure that is suitable for tractable analysis, seeking to mitigate queue latency and provide stability guarantees. We do this by analysing the model and leveraging classic control theory. The controller is evaluated both in a self-built Julia simulation framework and a state-of-the-art network simulator where it also is compared to state-of-the-art congestion control algorithms. The results are promising and show that the proposed control structure performs on par with the compared congestion controllers despite its simple nature. The most prominent area of improvement is in signal estimation, where we believe that the proposed controller could gain in performance and yield better results.