From Reactive to Agentic Resilience: Multi-Timescale AI-RAN Orchestration Under Signaling Storms
Organizations: 5G/6GIC, Institute for Communication Systems (ICS), University of Surrey, Guildford, UK
Abstract
Next-generation open radio access network (Open RAN) and AI-RAN deployments turn the RAN into a disaggregated, compute-aware platform whose service continuity depends on many control loops acting at different timescales. Resilience in this setting is best seen as a lifecycle spanning anticipation, absorption, adaptation, recovery, and evolution. Yet, existing mechanisms remain stage-specific and are rarely coordinated across the control hierarchy. This paper argues that agentic AI offers an orchestration layer that coordinates this lifecycle, turning operator intent into coordinated action while keeping slow language-model reasoning off the fast control path. Using mass-attachment signaling storms as a running study, we build a discrete-event Open RAN attach simulator that captures retry amplification, shared-compute contention, and non-instantaneous provisioning, and we propose a three-tier framework in which an orchestrator at the service management and orchestration layer sets a standing policy, a per-site non-real-time agent judges storms and anticipates scheduled crowds, and a deterministic near-real-time loop enforces every decision. Across experiments, the agentic controller approaches the resilience of a fully provisioned pool of servers at a quarter of its capacity by reading a scheduled event from a calendar and provisioning ahead of it. No reactive weight setting recovers this once the surge arrives. Free-text operator intents ground into control directives at 94% exact match on previously unseen intents, and a single instruction reshapes live behavior end to end. We also quantify the latency, cost, and contention limits that bound it. The supplementary material provides analytical proof of the derived Theorem on anticipation horizon, as well as the subsequent Corollary results on reactive provisioning limits.
Figures & tables
| Model | Overall Resil. (%) | Bot Resil. (%) | Surge Resil. (%) | Benign Attached (%) | Crowd Est. | Reserve Servers | Bot Filtered (%) | Mean Capacity | Latency/assess. | Cost/assess. |
| ( , true 83.7) | (srv, ideal 13) | (srv) | (s) | (\downarrow$ | ||||||
| GPT-5.4-Mini (on) | ||||||||||
| QWEN3.7-Plus | ||||||||||
| GEMINI-3.1-Flash-Lite | ||||||||||
| GPT-5.4-Mini (off) | ||||||||||
| CLAUDE-HAIKU-4.5 |
| System | Overall Resil. (%) | Capacity Efficiency |
| ( : full pool) | ||
| Static ( ) | ||
| Static ( ) | ||
| Static ( ) | ||
| Lyapunov | ||
| Rule-based |
| Event | Ideal | GPT-5.4-Mini | GEMINI-3.1-Flash-Lite |
| Sold out | |||
| Taylor Swift | 13 | 14.0 ( ) | 14.0 ( ) |
| Man U v Liverpool | 11 | 11.0 ( ) | 11.0 ( ) |
| Arsenal W v Man U W | 9 | 9.0 ( ) | 9.0 ( ) |
| Liverpool v Real Madrid | 8 | 8.9 ( ) | 8.7 ( ) |
| Adele | 3 | 3.0 ( ) | 3.0 ( ) |
| Overall (%) | Per-lever accuracy (%) | ||||||
| Model | Grounding | Exact match | Posture | Strength | Floor | Event | Delegation |
| GPT-5.4-Mini | 98.4 | 94.2 | 98 | 100 | 99 | 100 | 96 |
| GEMINI-3.1-Flash-Lite | 97.1 | 87.5 | 93 | 92 | 100 | 97 | 100 |
| Keyword matcher | 84.3 | 52.5 | 65 | 25 | 95 | 100 | 95 |
| Do-nothing | 66.3 | 5.0 | 35 | 0 | 82 | 85 | 82 |
| Metric | Baseline | +Intent | |
| Benign attached rate, (%) | 26.0 | 97.0 | |
| Benign false positive rate, (%) | 16.0 | 4.0 | |
| Resilience, (%) | 74.0 | 92.0 | |
| Mean capacity, (srv) | 3.05 | 3.57 |
| Controller | Resilience (%) | (pp) | ||
| Static ( ) | ||||
| Lyapunov | ||||
| Rule-based | ||||
| Agentic (GEMINI-3.1-Flash-Lite) | ||||
| Agentic (GPT-5.4-Mini) | ||||
| Site | Intent applied | (%) | (%) | (srv) |
| Stadium | QoS + legitimate (local override) | |||
| Residential | Fleet cost (not exempted) | |||
| Business | Cost + server floor |
| Sotiris Chatzimiltis (Graduate Student Member) received a B.Sc. degree in computer science from the University of Cyprus, Cyprus, in 2021, and an M.Sc. degree in computer vision, machine learning and robotics from the University of Surrey, U.K., in 2022. He is a PhD student at the Institute for Communication Systems at the University of Surrey. His current research interests include machine learning, intrusion detection systems, distributed AI, and Open RAN security. |
| Mahdi Boloursaz Mashhadi (S’14-M’18, SM’23) is a Lecturer at the 5G/6G Innovation Centre (5G/6GIC) at the Institute for Communication Systems (ICS), University of Surrey (UoS), and a Surrey AI fellow. His research is focused at the intersection of AI/ML with wireless communication, learning and communication co-design, generative AI for telecommunications, and collaborative machine learning. He received B.S., M.S., and Ph.D. degrees in mobile telecommunications from the Sharif University of Technology (SUT), Tehran, Iran. He has more than 40 peer reviewed publications and patents in the areas of wireless communications, machine learning, and signal processing. He is a PI/Co-PI for various government and industry funded projects including the UKTIN/DSIT 12M£ national project TUDOR. |
| Mohammad Shojafar (M’17-SM’19) is an Associate Professor in network security and an Intel Innovator, professional ACM member and ACM distinguished speaker, a fellow of the Higher Education Academy, and a Marie Curie alumnus, working in the 6G Innovation Centre (6GIC), Institute for Communication Systems (ICS), at the University of Surrey, UK. Before joining 6GIC, he was a senior researcher and a Marie Curie fellow in the SPRITZ Security and Privacy Research group at the University of Padua, Italy. Dr Shojafar secured around 3M as PI in various EU/UK projects. https://www.surrey.ac.uk/people/mohammad-shojafar . |
| Rahim Tafazolli (Fellow, IEEE) is currently a Professor in mobile and personal communications and the Director of the Institute of Communication Systems (ICS) and the 5G and 6G Innovation Centre, University of Surrey. He has been active in research for over 30 years and published more than 1200 research papers. He has been a technical advisor to many mobile companies, and has lectured, chaired, and been invited as keynote speaker to a number of IEEE workshops and conferences. He served as the Chairperson for the EU Expert Group on Mobile Platform (e-mobility SRA), the Chairperson for the Post-IP working Group in e-mobility, and the past Chairperson of WG3 of WWRF. He is nationally and internationally known in the field of mobile communications. In May 2018, he was appointed as a Regius Professor in electronic engineering for recognition of his exceptional contributions to digital communications technologies over the past 30 years. He was elected as a fellow of the U.K. Royal Academy of Engineering, in 2020. He is a fellow of IET and the Wireless World Research Forum (WWRF). |
Appendix figures & tables2 assets
Supplementary material from the paper’s appendix.
Appendix
| Symbol | Description | Value |
| Section II-A Scenario, Threat Model, and Service Rate | ||
| Benign false-positive fraction | ||
| Maximum servers | ||
| One-way transport delay | ms | |
| Attach arrival rate | — | |
| Benign / malicious arrivals | — | |
| Symbol | Description | Value |
| Section II-D Controllers | ||
| Normalized unmet demand | — | |
| Drift-plus-penalty objective | — | |
| Malicious drop probability | ||
| Normalized backlog | — | |
| Utility weight | ||
Open Radio Access Network
Open Radio Access Network (O-RAN) research studies disaggregated, programmable wireless networks and the intelligence that manages them.
See all 35 papers