This work introduces a standardized and lightweight abstraction framework for Deep Audio Embedding Neural Networks. Designed with edge and embedded scenarios in mind, the framework enables seamless integration, inference, and performance profiling of a large set of pre-trained Audio NNs, across diverse downstream domains and platforms, by introducing a unified API and an open-contributing configuration schema. This interface allows each model to be instantiated and executed through an easy interpretable descriptor, without altering the original developers source code. The proposed framework supports fully automated benchmarking orchestration in a cloud-based environment, with the aim of a modular infrastructure built on Amazon Web Services, which processes community-contributed model configurations and evaluates them on server-grade hardware. Benchmark profiling includes inference (CPU/GPU) timing, energy efficiency, and memory usage monitoring, minimizing runtime overhead and enabling fair comparison across implementations: results are aggregated into a real-time leaderboard and within technical model cards, accessible through a dedicated web interface. This initiative lays the groundwork for a community-driven benchmarking challenge, encouraging experimental reproducibility, scalability, and models cross-compatibility for real-world audio intelligence applications.

Urban Air Mobility (UAM) represents an innovative mobility solution designed to alleviate the traffic congestion associated with conventional two-dimensional transportation, which has already saturated both urban and underground spaces. The development of unmanned aerial vehicles (UAVs) and electric vertical takeoff and landing (eVTOL) aircraft is a key enabler for this new, sustainable mode of transporting goods and passengers. To ensure safe operation—especially as these vehicles advance toward autonomous capabilities—and to maintain connectivity for passengers, UAM requires high data-rate, reliable, and low-latency communications. Within the 6G framework, several architectural solutions have been proposed, incorporating components such as terrestrial stations, satellites, and UAVs. In this paper, we propose a two-tier architecture: an aerial tier that leverages multi-hop THz communications among UAM vehicles as backhaul links, and a terrestrial tier comprising conventional cellular base stations (BSs) along with a number of enhanced base stations (eBSs) equipped with high-elevation antenna systems to provide aerial coverage for UAM vehicles (UVs). We will show that our approach ensures reliable UAM connectivity even in high-density urban airspaces, striking an optimal balance between the required number of eBSs and the airborne THz backhauling.

Ensuring seamless connectivity for first responders represents a critical issue in emergency scenarios, where disrupted infrastructures and network congestion can significantly hinder the communication. Open Radio Access Network (O-RAN) promises a level of openness in Radio Access Networks (RANs) that could be advantageous under such critical conditions, allowing for enhanced integration across different operators’ networks to mitigate resource scarcity. This paper introduces an architectural solution based on O-RAN that leverages neutral hosting to provide a unified gateway to RAN infrastructures. Our approach facilitates seamless access for emergency responders and optimizes the allocation of resources at both the RAN and Core levels within a specified area by exploiting the deployment of near-real-time Applications (xApps) and non-real-time Applications (rApps) managed by the neutral host. The results demonstrate the effectiveness of O-RAN neutral hosting in enhancing first responders’ connectivity, showcasing significant improvements in throughput and latency, especially under disaster scenarios. • O-RAN Neutral Hosting is a feasible solution for emergency operator connectivity. • First responders seamlessly communicate through efficient radio resource allocation. • O-RAN Neutral Hosting outperforms traditional RAN before and after the disaster. • Slice isolation strategy further amplifies the advantages of O-RAN Neutral Hosting.

This paper introduces to a structured application of the One-Class approach and the One-Class-One-Network model for supervised classification tasks, specifically addressing a vowel phonemes classification case study within the Automatic Speech Recognition research field. Through pseudo-Neural Architecture Search and Hyper-Parameters Tuning experiments conducted with an informed grid-search methodology, we achieve classification accuracy comparable to nowadays complex architectures (90.0 - 93.7%). Despite its simplicity, our model prioritizes generalization of language context and distributed applicability, supported by relevant statistical and performance metrics. The experiments code is openly available at our GitHub.

With the advent of advanced communication technologies like 5G and the emergence of the Internet of Everything, there is a growing potential to leverage public communication networks for mission-critical applications. Traditionally, these applications have depended on private, narrowband networks such as TETRA and P25. The latest communication protocols and architectures promise specific levels of latency and reliability even during the negotiation phase. Concurrently, the increasing deployment of sensors for Structural Health Monitoring (SHM) offers access to vast data volumes at low costs and near real-time capabilities when required. This paper highlights the significant benefits of integrating these two paradigms in the context of Earthquake Early Warning (EEW) systems. Utilizing existing earthquake propagation models and recorded seismic data, we have developed a new simulator tailored to mimic a densely packed sensor network environment. We introduce a consensus algorithm that processes the initial milliseconds of data collected by sensors nearest to the earthquake’s hypocenter. An alert is then promptly issued. Our results indicate substantial improvements in reducing the time to relay warnings, thereby significantly diminishing the extent of the blind zone, where warnings arrive too late, in EEW systems.

The OPTIMIST project is positioned into one of the most challenging and societal impactful innovation paths, by contributing to the design and implementation of a Multi-access Edge Computing-empowered service platform enabling the optimized delivery of the most popular type of traffic in the forthcoming 5G networks: mobile video. Advanced prediction tools (e.g. data analytics), MEC-enabled service provisioning, 5G networking, QoE-driven and model-based service optimization using stochastic modeling and algorithmic optimization, are major components toward a holistic system that combines analytical tools and close-to-market technologies. In addition, despite the considerable amount of valuable academic research conducted in the area of MEC- enabled service provisioning, the vast majority of obtained results remains hardly compatible (in terms of system models and implementation platforms) and are often limited to the scope of academic research with no envisaged market transfer plans. OPTIMIST fosters fruitful collaboration of academia and industry targeting at developing an innovative platform for fully personalized delivery of 5G mobile video content.

Cyber-Physical Systems (CPS) are dynamic and reactive systems interacting with processes, environment and, sometimes, humans. They are often distributed with sensors and actuators, smart, adaptive, predictive and react in real-time. Indeed, as sight for human beings, image- and video-processing pipelines are a prime source for environmental information for systems allowing them to take better decisions according to what they see. Therefore, in FitOptiVis we are developing novel methods and tools to integrate complex image and video processing pipelines. FitOptiVis aims to deliver a reference architecture for describing and optimizing quality and resource management for imaging and video pipelines in CPS both at design- and run-time. The architecture is concretized in low-power, high-performance, smart components, and in methods and tools for combined design-time and run-time multi-objective optimization and adaptation within system and environment constraints.

The challenges in analysis and co-design of wireless networked control systems are well highlighted by considering wireless industrial control protocols. In this perspective, this paper addresses the modeling and design challenge by focusing on WirelessHART, which is a networking protocol stack widely adopted for wireless industrial automation. Specifically, we first develop and validate a Markov channel model that abstracts the WirelessHART radio link subject to channel impairments and interference. The link quality metrics introduced in the theoretical framework are validated in order to enable the accurate representation of the average and extreme behavior of the radio link. By adopting these metrics, it is straightforward to handle a consistent finite-state abstraction. On the basis of such a model, we then derive a stationary Markov jump linear system model that captures the dynamics of a control loop closed over the radio link. Subsequently, we show that our modeling framework is able to discover and manage the challenging subtleties arising from bursty behavior. A relevant theoretical outcome consists in designing a controller that guarantees stability and improves control performance of the closed-loop system, where other approaches based on a simplified channel model fail.

Cultural Heritage represents a resource of inestimable value in Italy. This work presents a project to allow people affected by specific mental disabilities to enjoy the experience of an exhibition, by exploiting experienced AAL technological solutions. Basing on the results obtained by Casa+, a project to support daily life of people affected by the Down syndrome, we designed and developed a system based on a flexible architecture providing a set of context-aware applications for assisting people with cognitive disabilities to experiment an autonomous cultural experience. This is achieved allowing a safe location reaching through outdoor navigation and an assisted personal experience inside the exhibition building. In particular, we propose a location-aware mobile navigation system that integrates a consecutive point navigation system for outdoor navigation and an indoor navigation system that integrates recommendation mechanisms based on the localization information. A case study for a real single smart space and some preliminary experimental results are presented and discussed.

The demo we propose represents a first step toward RF Sounding [1]. Such a project is an open space installation which comprises both artistic and technological innovations; its aim is to provide the user, while entering a specifically defined area, with awareness of radio frequency signals characterizing the cellular networks band. Indeed, radio signals are shifted, with proper elaboration, to the audible band and the result is spread all over the specific area through a certain number of loudspeakers. For this procedure we have been inspired by the eternal metaphor of the impossible human dream to fly, thus the limitations of our senses which are capable of feeling audio waves (sounds) but (maybe luckily) are not capable of directly feeling radio frequency (RF) waves are highlighted. The translation is the starting point for a more complex and exciting musical composition. It has to be underlined that some features of RF sensed signals are kept unchanged in order to give to the listener the clear feeling of the amount (power) of RF signals present in the monitored area. The aim of this project is twofold. Indeed, from one side we want to increase end users knowledge of the strength of the power emitted by their cellular phones with respect to the electromagnetic fields produced in the environment, on the other hand we want to provide for an artistic and interactive installation that can also be remotely joined through a web interface.

This paper focuses on the analysis of fundamental issues which has to be considered in order to gain voice from a wireless sensor network (WSN) that is assumed to be constituted by MicaZ wireless sensor nodes. Problems related to the transport of an audio signal through a wireless channel and sensor nodes are thus analyzed and a project for an audio surveillance system is presented.

An adaptive handover algorithm for wireless communication systems is addressed in this paper. Moving from the Generalized Extended Least Square handover algorithm proposed in N. Benvenuto et al. (2002), we model the handover mechanism as a hybrid system, and we include it in a dynamic optimization problem which is solved through the use of a trellis diagram. The objective function takes into account the parameters that influence the handover mechanism. The optimal solution of the program allows to evaluate performance of the handover in terms of the probability of handover and probability of outage.

This paper proposes an analytical framework for performance evaluation of TCP (transport control protocol) over adaptive wireless links. Specifically, we include adaptation of power, modulation format and error recovery strategy, and incorporate some features of wireless fading channels. This framework is then used to pursue joint optimization through maximization of an objective functional, that expresses a trade-off between achievable throughput and energy costs. A set of numerical results is reported, and it is seen that hybrid ARQ schemes may provide significant benefits in the optimization framework