Candidate: Anas Abognah
Date: March 24, 2026
Time: 10:00 AM
Location: Online (Microsoft Teams)
Supervisor: Otman Basir
Abstract:
As radio frequency spectrum becomes increasingly scarce, the telecommunications industry is moving toward Dynamic Spectrum Management (DSM). However, current approaches rely on centralized brokers that introduce single points of failure and privacy risks, while existing AI solutions lack the high-level strategic reasoning required for complex markets.
This seminar introduces BLAST, a novel framework that integrates Large Language Model (LLM) Agents with a permissioned Hyperledger Fabric blockchain to create a fully autonomous, private, and secure spectrum trading ecosystem. We will discuss the system's unique architecture, which utilizes a "Perceive-Plan-Act" cognitive cycle for agents, and present game-theoretic evaluations of various auction mechanisms.
Experimental results demonstrate that the Second-Price (Vickrey) auction maximizes social welfare (capturing up to 71% of theoretical surplus) by incentivizing truthful bidding. Furthermore, we show that LLM-driven agents outperform traditional heuristic baselines in complex market scenarios, achieving higher allocative efficiency and lower wealth inequality (Gini coefficient). This work highlights the potential of Generative AI to serve as the cognitive core for next-generation decentralized wireless networks.