arXiv subjects

Kleber Cardoso

Publications and source records attributed to Kleber Cardoso.

5 recordsLinked to original sources

arXiv2026

Enabling Real-Time Programmability for RAN Functions: A Wasm-Based Approach for Robust and High-Performance dApps

While the Open Radio Access Network Alliance (O-RAN) architecture enables third-party applications to optimize radio access networks at multiple timescales, real-time distributed applications (dApps) that demand low latency, high performance, and strong isolation remain underexplored. Existing approaches propose colocating a new RAN Intelligent Controller (RIC) at the edge, or deploying dApps in bare metal along with RAN functions. While the former approach increases network complexity and requires additional edge computing resources, the latter raises serious security concerns due to the lack of native mechanisms to isolate dApps and RAN functions. Meanwhile, WebAssembly (Wasm) has emerged as a lightweight, fast technology for robust execution of external, untrusted code. In this work, we propose a new approach to executing dApps using Wasm to isolate applications in real-time in O-RAN. Results show that our lightweight and robust approach ensures predictable, deterministic performance, strong isolation, and low latency, enabling real-time control loops.

cs.NI↗

arXiv2025

Towards a Robust Transport Network With Self-adaptive Network Digital Twin

The ability of the Network digital twin (NDT) to remain aware of changes in its physical counterpart, known as the physical twin (PTwin), is a fundamental condition to enable timely synchronization, also referred to as twinning. In this way, considering a transport network, a key requirement is to handle unexpected traffic variability and dynamically adapt to maintain optimal performance in the associated virtual model, known as the virtual twin (VTwin). In this context, we propose a self-adaptive implementation of a novel NDT architecture designed to provide accurate delay predictions, even under fluctuating traffic conditions. This architecture addresses an essential challenge, underexplored in the literature: improving the resilience of data-driven NDT platforms against traffic variability and improving synchronization between the VTwin and its physical counterpart. Therefore, the contributions of this article rely on NDT lifecycle by focusing on the operational phase, where telemetry modules are used to monitor incoming traffic, and concept drift detection techniques guide retraining decisions aimed at updating and redeploying the VTwin when necessary. We validate our architecture with a network management use case, across various emulated network topologies, and diverse traffic patterns to demonstrate its effectiveness in preserving acceptable performance and predicting quality of service (QoS) metrics under unexpected traffic variation, such as delay and jitter. The results in all tested topologies, using the normalized mean square error as the evaluation metric, demonstrate that our proposed architecture, after a traffic concept drift, achieves a performance improvement in per-flow delay and jitter prediction of at least 64% and 21%, respectively, compared to a configuration without NDT synchronization.

cs.NI↗

arXiv2025

Network Digital Twin for Route Optimization in 5G/B5G Transport Slicing with What-If Analysis

The advent of fifth-generation (5G) and Beyond 5G (B5G) networks introduces diverse service requirements, from ultra-low latency to high bandwidth, demanding dynamic monitoring and advanced solutions to ensure Quality of Service (QoS). The transport network - responsible for interconnecting the radio access network and core networks - will increasingly face challenges in efficiently managing complex traffic patterns. The Network Digital Twin (NDT) concept emerges as a promising solution for testing configurations and algorithms in a virtual network before real-world deployment. In this context, this work designs an experimental platform with NDT in a transport network domain, synchronizing with the virtual counterpart and a recommendation system for what-if analysis, enabling intelligent decision-making for dynamic route optimization problems in 5G/B5G scenarios. Our NDT, composed of a Graph Neural Network (GNN), was evaluated across three different network topologies consisting of 8, 16, and 30 nodes. It achieved lower MAPE values for URLLC and eMBB slices, comparing latency predictions with actual latency after the solution implementation. These values indicate high accuracy, demonstrating the solution's effectiveness in generating precise insights into network performance if a particular solution were implemented.

cs.NI↗

arXiv2025

Intent-based Radio Scheduler for RAN Slicing: Learning to deal with different network scenarios

The future mobile network has the complex mission of distributing available radio resources among various applications with different requirements. The radio access network slicing enables the creation of different logical networks by isolating and using dedicated resources for each group of applications. In this scenario, the radio resource scheduling (RRS) is responsible for distributing the radio resources available among the slices to fulfill their service-level agreement (SLA) requirements, prioritizing critical slices while minimizing the number of intent violations. Moreover, ensuring that the RRS can deal with a high diversity of network scenarios is essential. Several recent papers present advances in machine learning-based RRS. However, the scenarios and slice variety are restricted, which inhibits solid conclusions about the generalization capabilities of the models after deployment in real networks. This paper proposes an intent-based RRS using multi-agent reinforcement learning in a radio access network (RAN) slicing context. The proposed method protects high-priority slices when the available radio resources cannot fulfill all the slices. It uses transfer learning to reduce the number of training steps required. The proposed method and baselines are evaluated in different network scenarios that comprehend combinations of different slice types, channel trajectories, number of active slices and users' equipment (UEs), and UE characteristics. The proposed method outperformed the baselines in protecting slices with higher priority, obtaining an improvement of 40% and, when considering all the slices, obtaining an improvement of 20% in relation to the baselines. The results show that by using transfer learning, the required number of training steps could be reduced by a factor of eight without hurting performance.

cs.NI↗

arXiv2021

System Intelligence for UAV-Based Mission Critical with Challenging 5G/B5G Connectivity

Unmanned aerial vehicles (UAVs) and communication systems are fundamental elements in Mission Critical services, such as search and rescue. In this article, we introduce an architecture for managing and orchestrating 5G and beyond networks that operate over a heterogeneous infrastructure with UAVs' aid. UAVs are used for collecting and processing data, as well as improving communications. The proposed System Intelligence (SI) architecture was designed to comply with recent standardization works, especially the ETSI Experiential Networked Intelligence specifications. Another contribution of this article is an evaluation using a testbed based on a virtualized non-standalone 5G core and a 4G Radio Access Network (RAN) implemented with open-source software. The experimental results indicate, for instance, that SI can substantially improve the latency of UAV-based services by splitting deep neural networks between UAV and edge or cloud equipment. Other experiments explore the slicing of RAN resources and efficient placement of virtual network functions to assess the benefits of incorporating intelligence in UAV-based mission-critical services.

cs.NI↗