AI Lifecycle-Aware Feasibility Framework for Split-RIC Orchestration in NTN O-RAN

Daniele Tarchi

cs.NIMar 24, 2026

AI Lifecycle-Aware Feasibility Framework for Split-RIC Orchestration in NTN O-RAN

Authors: Daniele Tarchi

Organizations: Department of Information Engineering, University of Florence, 50139 Firenze, Italy · CNIT - University of Florence Research Unit, 50139 Firenze, Italy

Abstract

Integrating Artificial Intelligence (AI) into Non-Terrestrial Networks (NTN) is constrained by the joint limits of satellite SWaP and feeder-link capacity, which directly impact O-RAN closed-loop control and model lifecycle management. This paper studies the feasibility of distributing the O-RAN control hierarchy across Ground, LEO, and GEO segments through a Split-RIC architecture. We compare three deployment scenarios: (i) ground-centric control with telemetry streaming, (ii) ground--LEO Split-RIC with on-board inference and store-and-forward learning, and (iii) GEO--LEO multi-layer control enabled by inter-satellite links. For each scenario, we derive closed-form expressions for lifecycle energy and lifecycle latency that account for training-data transfer, model dissemination, and near-real-time inference. Numerical sensitivity analysis over feeder-link conditions, model complexity, and orbital intermittency yields operator-relevant feasibility regions that delineate when on-board inference and non-terrestrial learning loops are physically preferable to terrestrial offloading.

Figures & tables

Param.	Ground ( $\mathcal{G}_{st}$ )	LEO ( $\mathcal{S}_{LEO}$ )	GEO ( $\mathcal{S}_{GEO}$ )
HW Type	Cloud GPU	Rad-Hard FPGA	Space DSP
Capacity ( $F_{n}$ )	$>$ 1\text{,}\mathrm{P}$$	$\approx$ 10\text{,}\mathrm{T}$$	$\approx$ 200\text{,}\mathrm{T}$$
Power ( $P_{n}^{max}$ )	Unlimited	$20\text{\,}\mathrm{W}$	$1000\text{\,}\mathrm{W}$
Eff. ( $\epsilon_{n}$ )	$200\text{\,}\mathrm{pJ}$ /FLOP	$20\text{\,}\mathrm{pJ}$ /FLOP	$100\text{\,}\mathrm{pJ}$ /FLOP

TABLE I: Node Resource Specifications

Domain	Assumption	Justification
Data	Raw Telemetry	We assume O1 data ( $\delta^{in}$ ) is uncompressed to model the worst-case bandwidth load of spectral monitoring (e.g., I/Q samples).
Compute	Hardware Heterogeneity	LEO nodes use energy-efficient INT8 quantization ( $\epsilon_{LEO}$ ), while Ground/GEO nodes use high-precision FP32 ( $\epsilon_{GND}$ ), reflecting standard edge-vs-cloud AI practices.
Power	Thermal Throttling	We assume the LEO power budget ( $P_{comp}$ ) is an average value that accounts for duty-cycling required to prevent thermal overheating.
Topology	ISL Availability	The GEO-LEO optical link is assumed to be persistently available during the update window, treating the GEO satellite as a “Data Relay” backbone.
Training	One-Shot Learning	We model the training phase as a discrete event ( $T_{train}$ ); iterative Federated Learning rounds are aggregated into a single energy cost for clarity.

TABLE II: Summary of Modeling Assumptions

Panel of Figure 4 — Fig. 2: Comparative O-RAN Deployment Architectures. (a) S1: Traditional Bent-Pipe approach with all intelligence grounded. (b) S2: Proposed Split-RIC with inference at the edge and training on the ground. (c) S3: Hierarchical approach with inference at the edge and training on a GEO hub via ISL.

Parameter	Symbol	Value
Orbital & Link Parameters
Orbit Altitude	$h_{LEO}$	$600\text{\,}\mathrm{km}$
Feeder Link Frequency	$f_{c}$	$28\text{\,}\mathrm{GHz}$ (Ka-band)
Feeder Link Bandwidth	$B_{feed}$	$400\text{\,}\mathrm{MHz}$
Feeder Tx Power	$P_{tx}^{feed}$	$15\text{\,}\mathrm{W}$
Feeder Rx Power	$P_{tx}^{feed}$	$5\text{\,}\mathrm{W}$

TABLE III: Simulation Parameters

Fig. 3: Energy Feasibility Region: Impact of Data Volume

Fig. 5: Energy Feasibility Region: Impact of Complexity

Fig. 7: Latency Feasibility: Impact of Intermittency

Open Radio Access Network

Open Radio Access Network (O-RAN) research studies disaggregated, programmable wireless networks and the intelligence that manages them.

See all 28 papers

Toward Autonomous O-RAN: A Multi-Scale Agentic AI Framework for Real-Time Network Control and Management
Feb 15, 2026Hojjat Navidan, Mohammad Cheraghinia, Jaron Fontaine +5Multi-Agent Orchestration Language Model-Based Control
IDLab, Department of Information Technology at Ghent University - imec, Ghent, Belgium · Department of Electrical and Computer Engineering, Princeton University, USA
X-RIC: An Agentic RAN-Intelligent Controller Benchmark for Enhancing Reinforcement Learning
Oct 15, 2026Adhwaa Al-Chaab, Ayman Younis, Dario PompiliNetwork Slicing Agentic RL
†Department of Electrical and Computer Engineering, Rutgers University–New Brunswick, NJ, USA
Agentic AI-RAN: Enabling Intent-Driven, Explainable and Self-Evolving Open RAN Intelligence
Feb 27, 2026Zhizhou He, Yang Luo, Xinkai Liu +4Network Slicing Wireless Communications
5G/6GIC, Institute for Communication Systems (ICS), University of Surrey, Guildford, UK · 6G Research Center, Khalifa University, Abu Dhabi, UAE

Feb 15, 2026J/K move · Enter open · S save

Toward Autonomous O-RAN: A Multi-Scale Agentic AI Framework for Real-Time Network Control and Management

Hojjat Navidan, Mohammad Cheraghinia, Jaron Fontaine, Mohamed Seif+4

IDLab, Department of Information Technology at Ghent University - imec, Ghent, Belgium · Department of Electrical and Computer Engineering, Princeton University, USA

Multi-Agent Orchestration Language Model-Based Control Network Slicing Wireless Communications Open Radio Access Network Intent-Based Networking+1

Open paper AskPDF arXiv

Open Radio Access Networks (O-RAN) promise flexible 6G network access through disaggregated, software-driven components and open interfaces, but this programmability also increases operational complexity. Multiple control loops coexist across the service management layer and RAN Intelligent Controller (RIC), while independently developed control applications can interact in unintended ways. In parallel, recent advances in generative Artificial Intelligence (AI) are enabling a shift from isolated AI models toward agentic AI systems that can interpret goals, coordinate multiple models and control functions, and adapt their behavior over time. This article proposes a multi-scale agentic AI framework for O-RAN that organizes RAN intelligence as a coordinated hierarchy across the Non-Real-Time (Non-RT), Near-Real-Time (Near-RT), and Real-Time (RT) control loops: (i) A Large Language Model (LLM) agent in the Non-RT RIC translates operator intent into policies and governs model lifecycles. (ii) Small Language Model (SLM) agents in the Near-RT RIC execute low-latency optimization and can activate, tune, or disable existing control applications; and (iii) Wireless Physical-layer Foundation Model (WPFM) agents near the distributed unit provide fast inference close to the air interface. We describe how these agents cooperate through standardized O-RAN interfaces and telemetry. Using a proof-of-concept implementation built on open-source models, software, and datasets, we demonstrate the proposed agentic approach in two representative scenarios: robust operation under non-stationary conditions and intent-driven slice resource control.

Abstract

Open Radio Access Network

Explore similar work

Toward Autonomous O-RAN: A Multi-Scale Agentic AI Framework for Real-Time Network Control and Management

X-RIC: An Agentic RAN-Intelligent Controller Benchmark for Enhancing Reinforcement Learning

Agentic AI-RAN: Enabling Intent-Driven, Explainable and Self-Evolving Open RAN Intelligence