CAREER: CatFly: Towards Resilience-Native Wireless Networks through Learning, Twinning, and Reconfiguring Co-Design

The evolution of next-generation (NextG) mobile and wireless networks is driven by a move toward higher carrier frequencies, such as millimeter-wave (mmWave) bands. Higher frequencies provide higher capacity but also have a much shorter distance range for coverage compared to lower-frequency signals. This means that a single access point (AP) or a base station cannot cover a large area leading to smaller areas (cells) with each AP handling a smaller number of users. This dense deployment of small-cell APs necessitates a heightened level of intelligence and timely situational awareness to enhance network resilience and self-reconfigurability in the face of various challenges like network or AP failures. To tackle these challenges, this project pursues a novel resilience-native network paradigm called CatFly, which embraces a data-driven learning approach that utilizes a digital replica of the physical network with sufficient details to swiftly achieve preemptive operations against disruptions. Armed with this hybrid digital-physical (HDP) intelligence, networks are always ready and responsive, employing outcomes of their what-if analysis to reconfigure the physical network and ensure resilience. The project aims to utilize a combination of techniques such as network optimization, graph theory, machine learning, experimental measurements, and models that mix physical and virtual contexts. The project will advance networking technologies in three inter-related thrusts, followed by a comprehensive system-level analysis and validation, including: 1) Generalized prediction frameworks that learn critical performance indicators for spatio-temporal awareness and cross-domain knowledge, linking them to handle network disruptions; 2) Multi-scale approach to creating and evolving a network digital replica with sufficient details that is faithful to the physical network to assist with resilience-centric awareness; 3) Hybrid digital-physical optimization with a stabilization-driven mechanism to uncover multi-dimensional physical reconfigurations. The project will develop a comprehensive educational plan that includes a pre-college STEM virtual laboratory, new network intelligence-centered course materials, and hands-on activities using the designed software tools and testbeds. A project website that provides access to code, data, and educational related resources will be actively maintained and updated for the duration of the project. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.