Extreme Value Theory Converges with Generative AI for Ultra-Reliable Low Latency Communications

Purpose and goal

The EVTAI project aims to develop the theoretical foundations of generative AI and wireless communications to achieve ultra-reliable low-latency communication (URLLC) for critical applications like remote surgery and autonomous vehicles. Current systems fail to address rare extreme events impacting reliability and latency. EVTAI integrates Extreme Value Theory (EVT) with generative AI to estimate extreme events and meet URLLC requirements, ensuring real-time, reliable communication.

Expected effects and result

The project aims to advance generative AI and wireless communications, driving research in EVT and generative AI for URLLC systems. Additionally, the project will enhance our fundamental understanding of ultra-reliable and low-latency constraints through the validation of novel EVT algorithms, generative AI techniques, and wireless channel tail estimation tailored for these applications. Outcomes include at least five high-quality publications and the mentorship of multiple students.

Planned approach and implementation

We explore joint modeling of reliability and latency using MEVT and wireless communication, expanding EVT and wireless methods with machine learning for real-time URLLC estimation. By integrating generative AI into statistical models, we aim to improve accuracy, address real-time demands, and ensure low latency and high reliability. Experimental data from intra-vehicular communication will validate these objectives, demonstrating enhanced learning accuracy with and without our methods.