PROGRAMMA PRIN 2022 - BCALL PRIN2022 D.D. N.104 DEL 2 FEBRUARY 2022
PROJECT TITLE: LOREN - LOw-delay congestion control for REal-time applications over the iNternet
CODICE CUP D53D23000960006
Budget: € 135.200
P.I. or U.R. Manager: Prof. Eng. Saverio Mascolo
Other Research Units or any Sub-Units: Sapienza University of Rome
Brief project description
Congestion control in packet networks has historically been aimed at preventing overload while ensuring high network utilization, and has been implemented through congestion control in TCP.
Today, the growing spread of real-time applications - such as videoconferencing, virtual reality, remote control even from LEO satellites - also requires latency control. In this context, the project aims to develop an algorithm capable of controlling both congestion and transmission delays.
The proposed approach is based on control theory to design solutions with formally provable stability and efficiency properties. The algorithm is applicable both to wired networks and to 5G scenarios characterized by high variability and intermittent connectivity, where traditional TCP protocols are less effective. While respecting the end-to-end principle, advanced network functionalities such as SDN, cross-layer optimizations, and in-band telemetry are also considered.
Purpose
The purpose of the project is to overcome the limitations of traditional TCP by using control theory for a new low-delay protocol. The project also aims to integrate control at different levels of the network stack, from transport to the application layer, addressing issues such as synchronization in distributed systems, dynamic resource adaptation, and congestion control in vehicular communications.
Expected results
The expected results include the development of a rigorous theoretical framework for modeling and analyzing congestion and latency control systems for real-time traffic. Advanced simulations and tests on real networks, including production networks and 5G testbeds, are expected to demonstrate superior performance in terms of congestion, channel utilization, and latency.
From an application perspective, the project aims to significantly improve the quality and reliability of communications for time-sensitive applications, with important scientific, technological, and socio-economic impacts.
Achieved results
The QUIC-DC (QUIC delay control) algorithm has been designed and implemented, capable of exploiting one-way delay variation to anticipate congestion events and thereby control queuing delay. Experimental evaluations, in controlled and real environments (including LEO satellite networks), show a significant reduction in latency and losses, while maintaining efficiency and fairness.
Relevant results have also been achieved in streaming: a distributed playback synchronization strategy based on consensus and event-triggered control, with reduced communication overhead, and a low-complexity Adaptive Bitrate algorithm based on Model Predictive Control, with improvements in Quality of Experience. In the 5G NR-V2X context, models and algorithms for congestion control oriented toward the Age of Information have been developed, improving data freshness and reliability.
