repositories

LLM-powered Agentic 6G Cross-Domain Negotiation Github Repo

This project simulates an A2A-like negotiation between two LLM-powered agents, a RAN (Energy Saving) Agent and an Edge (Latency Assurance) Agent, to reconcile their conflicting goals. They use a small-scale digital twin (DT) to validate their proposals/counter-proposals before sending them to the peer agent. The simulation evaluates different strategies, including using a collective memory with and without debiasing mechanisms.

LLM-powered Agentic 6G Network Slicing Negotiation Github Repo

A Python simulation environment where autonomous LLM-powered agents, representing different network slices (e.g., eMBB and URLLC) negotiate for RAN bandwidth and Edge CPU resources to meet their respective SLAs while minimizing energy consumption. The agents invoke their small-scale predictive Digital Twins to validate their proposals/counter-proposals.

This project explores how different reasoning strategies — Neglect of Uncertainty (mean-based) vs. Risk-Aware Mitigation (CVaR-based) — impact network performance, SLA violations, and energy efficiency.

mfoxh-GPU

A GPU-enabled Matlab implementation of the multivariate Fox H-function. It can also be used on CPU by dropping gpuArray() function.

mfoxh-C-MEX

A C/MEX parallel multi-thread implementation for the multivariate Fox H-function.