Consider a team of autonomous robots coordinating in real time or a remote medical monitoring system streaming sensor data from thousands of devices. Each of these cyber-physical systems competes for scarce communication resources. When networks are flooded with irrelevant data, critical signals, such as sensor failures or urgent alarms, risk being overlooked. What if, instead of transmitting all available data, networks could prioritize and send only what matters, when it matters?
Modern networks were originally designed to transmit data, not to interpret its significance. Every bit, packet, and stream is treated identically, regardless of its criticality or relevance to the final goal. But as cyber-physical systems grow more complex and demanding, this indiscriminate approach is proving inadequate. A new philosophy is taking shape: rather than treating all data as equally important, networks can prioritize information relevant to a predefined goal, or « intent. » This intent-based, minimalist philosophy draws on a classic extension of Shannon’s information theory to improve communication effectiveness by allocating communication and computation resources more judiciously.
Intent must shape the network from the bottom up.
To understand why this is critical, consider a connected health monitoring system. A bradycardia warning that arrives a few milliseconds late is indistinguishable from a routine step-count report at the network level. The former constitutes a clinically unacceptable failure, while the latter can tolerate minutes of delay. A traditional network is blind to this distinction. An intent-based network is not: it knows what matters, to whom, and when.
Turning this vision into reality requires more than a clever algorithm. It requires a fundamental, full-stack redesign. A unified service management and orchestration layer must sit alongside the traditional protocol stack. It coordinates application-level functions such as feature extraction, compression, and coding, while also directing resource allocation in the Radio Access Network (RAN). In short, intent must shape the network from the bottom up, spanning Physical (PHY) and Medium Access Control (MAC) layer operations to high-level processing at the application layer.
This approach builds on the broader principles of Intent-Based Networking (IBN), which automates how networks profile, activate, and assure intents. However, conventional IBN typically operates at higher layers, such as service orchestration and network management, and responds on timescales of seconds or longer because it relies on centralized decision-making and broad network visibility. Semantics-aware communication pushes intent realization closer to where data is processed and transmitted, deeper into both the application layer and RAN. This enables the network to adapt within milliseconds in response to varying intents and network dynamics.
Building the Right Foundation
This is where Open RAN becomes essential. By introducing openness, disaggregation, and programmability into the RAN, Open RAN provides the foundation for integrating intent-based, semantics-aware functionality throughout the protocol stack. The O-RAN Alliance’s architecture positions the RAN Intelligent Controller (RIC) as a central platform for non-real-time (non-RT) and near-real-time (near-RT) control of RAN operations. However, the RIC’s control-loop latency limits how quickly it can respond to fast-changing intents.
To address this limitation, a newer concept has emerged: distributed applications (dApps) that move intelligence closer to the lower layers of the RAN. By shortening control loops, reducing signaling overhead, and granting granular access to fine-grained local network data such as per-packet information and raw I/Q samples, dApps unlock truly real-time, intent-based decision-making. However, dApps remain largely a research concept for now, as their integration into the Open RAN architecture is still under active discussion and not yet part of any finalized standard.
ib-ORAN supports metrics and KPIs that allow evaluation not only of throughput or latency but also of how well the network serves a specific intent.
From Blueprint to Benchmark: The ib-ORAN Framework
Realizing this vision, however, raises a practical question: how can the Open RAN architecture with intent support, comprising intent-based, semantics-aware dApps, rApps (Non-RT RIC applications), and xApps (Near-RT RIC applications), be tested and validated before real-world deployment? Answering this requires a simulation environment that is lightweight enough to prevent excessive processing overhead and realistic enough to reflect a close-to-real network environment, while complying with 3rd Generation Partnership Project (3GPP) and O-RAN Alliance standards and specifications, thereby reducing the gap between development and real-world implementation.
The Orange Research team has developed ib-ORAN, an open-source and reproducible simulation framework for realistic evaluation and benchmarking of intent-based orchestration in Open RAN. The ib-ORAN framework enables coordinated control across multiple timescales through dApps, xApps, and rApps while adhering to Open RAN architectural principles and 3GPP standards.
Various RAN applications, such as Radio Resource Management (RRM), multiuser scheduling, and load balancing, whether conventional or AI-driven, can be deployed seamlessly across RIC and local dApp components, depending on the design principles and required control timescales. The simulator also supports a range of traffic data formats, including text, images, and semantic embeddings.
ib-ORAN is released as the « ns3-iboran » module and is openly available on the Orange GitHub repository.
Under the Hood: How ib-ORAN Works
ib-ORAN is built on top of the ns-3 framework, a widely used open-source network simulator that provides a realistic, event-driven environment for modelling wireless and wired networks. Compliant with 3GPP standards, ns-3 supports the kind of detailed radio behavior, such as mobility models, propagation effects, and standard RAN procedures, that make simulation results meaningful and transferable to real-world deployment.
On top of this foundation, ib-ORAN integrates Python-based control modules that mirror the standard Open RAN hierarchy: dApps for local, fine-grained control, while xApps and rApps host higher-layer RIC operations. These modules connect to the ns-3 simulation core through a high-performance, low-latency interface, allowing the control logic to interact with the network environment in a realistic, standard-compliant way.
Beyond the control structure, ib-ORAN supports defining custom, intent-specific performance metrics alongside standard key performance indicators (KPIs), allowing researchers to evaluate network behavior not just in terms of throughput or latency, but in terms of how well the network serves a specific intent.
About the Project
ib-ORAN was developed as part of the ANR COMSEMA project, which targets video interpretation applications and proposes a fundamental intent-based, semantics-aware approach to redesigning the entire process of information generation and transmission across the network, with particular emphasis on multi-source scenarios.
Conclusion and Perspectives
ib-ORAN demonstrates that the question this article opened with, « what if networks only transmitted what truly matters? », is not merely a philosophical ambition. It is an engineering challenge that can be systematically studied, measured, and refined. From intent definition to real-time radio resource decisions, this simulation framework enables rigorous evaluation of intent-based orchestration across the control stack, before moving to real hardware deployment. The path ahead is long, but ib-ORAN marks a concrete step forward.
As the ib-ORAN framework supports a range of traffic data formats, it opens the door to more sophisticated applications, such as video interpretation, without requiring major changes to the simulator’s RAN components. Looking further ahead, the framework’s modular design naturally extends to support emerging AI use cases, for instance, scenarios in which AI agents communicate directly with each other, exchanging embeddings and semantic representations across the network. Semantic Agent-to-Agent communication introduces a new class of intent-aware traffic that ib-ORAN is well-positioned to support.
Glossary
Open RAN (Open Radio Access Network)
An open, disaggregated approach to building Radio Access Networks, where hardware and software from different vendors can work together through standardized interfaces, instead of relying on a single, closed vendor solution.
IBN (Intent-Based Networking)
A network management approach where operators define high-level intents, and the network automatically configures itself to meet them, rather than being manually configured step by step.
RIC (RAN Intelligent Controller)
The central platform in Open RAN, responsible for managing and optimizing network behavior through intelligent applications, operating at both near-real-time and non-real-time timescales.
dApp (Distributed Application)
A lightweight control application that runs close to the radio hardware, enabling very fast, local decision-making with access to fine-grained, real-time network data. Nonetheless, dApps are not yet part of any finalized O-RAN standard and remain an active area of research and specification.
xApp (Near-RT RIC Application)
An application that consumes and/or produces Near-RT RIC services via standardized APIs to provide near-real-time control of, or guidance to, RAN nodes, operating on timescales of milliseconds to seconds.
rApp (Non-RT RIC Application)
An application that consumes and/or produces non-real-time management and automation services within the Non-RT RIC, supporting longer-term policy and network optimization on the scale of seconds.
Read more
Intent-Based Orchestration in Open RAN: An ns-3 Simulation Framework
The original research paper that presents the ib-ORAN framework, covering its architecture, design principles, and a first use case implementing intent-based radio resource management with simulation results.
Semantic Communication for Future Networks: ANR COMSEMA
The ANR-funded research project behind ib-ORAN, which targets intent-based, semantics-aware communication for video interpretation and multi-source scenarios across future network architectures.
Compacting Semantic Matryoshka Representations with Product Quantization
A research paper exploring efficient compression of semantic representations using product quantization, which is directly relevant to the application-layer pre-processing capabilities envisioned in ib-ORAN.
Semantic Communication for People Re-Identification
A research paper applying semantics-aware communication to the people re-identification task, as one of the target applications driving the design of the ANR COMSEMA project and the ib-ORAN framework.
Authors
Postdoctoral Researcher at Orange
Senior Researcher at Orange
