Autonomous maintenance for satellite-terrestrial integrated networks

点击翻译按钮获取中文摘要

Reliable broadband Internet has never become so indispensable in our daily lives during and after the pandemic world. However, in 2021 alone, days of countrywide Internet outages have happened a few times. The Rogers network outage in April and social media service outage in October caused billion-dollar costs. The future Internet will face serious challenges. On the one hand, it is expected to provide global broadband access to more than 85% of the global population who currently have no broadband as reported by the International Telecommunications Union. On the other hand, highly resilient Internet access needs to be ensured with the increasing complexity of global telecommunications infrastructures in space and ground segments. With the new launch plans for massive non-geostationary (NGSO) satellites, the future Internet infrastructure will brace the convergence of satellite and terrestrial networks based on a new paradigm called satellite-terrestrial integrated network (STIN). A STIN is envisioned to enable global broadband Internet with the integration of advanced medium-Earth-orbit (MEO) and low-Earth-orbit (LEO) satellite constellations into heterogeneous ground segments. However, this vision can hardly be fulfilled without effective maintenance and operations for resilience assurance. For example, malfunctioning components, atmospheric events, or anomalous traffic on any network segment can easily degrade or disrupt network services. The complex, heterogeneous, and dynamic nature of the space and ground components introduces persistent operations challenges. Equipping STINs with an autonomous maintenance (AM) capability offers an ideal solution to the challenges. Such a solution can handle the increasing complexity of STINs and mitigate performance degradation. The long-term goal of the proposed research is to develop novel AM approaches to making the STIN-based Internet infrastructure resilient, intelligent, and self-maintainable. The short-term objectives are to explore the fundamental elements of AM in root cause analysis and anomaly mitigation for an NGSO-satellite-based STIN infrastructure. The proposed research will be conducted through holistic data-driven STIN architectures, generalizable anomaly detection models, and anomaly mitigation schemes with novel machine learning (ML) models and frameworks. The proposed research will advance the field by creating approaches and techniques that can be applied to autonomous management for any STIN-related network. The research will generate impact in academia and industry and ultimately boost the digital economy worldwide. It will also contribute to Canada's national priority for enhancing the digital infrastructure for unserved and underserved areas. HQP will be trained to conduct key research objectives and gain unique and transferrable skills sets of ML, data analytics, networking, and software development, making them highly competitive in securing positions in academia and industry.