Poster: VeriRAN: Explainable and Runtime-Verified Multi-Agent Control for Trustworthy AI-RAN
TL;DR
VeriRAN is a lightweight runtime-verified AI-RAN control architecture that separates intelligence from authorization, and uses AI agents as intelligent action proposers that generate radio-aware and service-level-agreement (SLA)-aware decisions.
Abstract
Artificial intelligence (AI)-Native sixth-generation (6G) radio access networks (RANs) are expected to support increasingly dynamic control loops, where learning-based agents continuously adapt radio resources, slice priorities, and service behavior. However, this shift from static optimization to autonomous control raises a fundamental trust question: how can AI-driven RAN decisions remain safe when radio conditions change, telemetry becomes unreliable, or latency-critical services experience sudden stress? This paper presents VeriRAN, a lightweight runtime-verified AI-RAN control architecture that separates intelligence from authorization. Instead of treating AI agents as trusted actuators, VeriRAN uses them as intelligent action proposers that generate radio-aware and service-level-agreement (SLA)-aware decisions. Our verification shield is placed directly in the control path to approve, constrain, or replace unsafe actions before they affect the RAN.
