Next-generation autonomous and networked industrial systems (i.e., robots, vehicles, drones) have driven advances in ultra-reliable low-laten-cy communications (URLLC) computing. These multi-agent require fast, communication-efficient, distributed machine learning (ML) to provide mission-crit-ical control functionalities. Distributed ML techniques, including federated (FL), represent a mushrooming multidisciplinary research area weaving together sensing, communication, learning. FL enables...

The advent of the fourth industrial revolution (Industry 4.0) aims at increasing automation and efficiency in manufacturing processes by adoption information communication technologies. Several proposed solutions rely on precise localization material, equipment, or operators. This article investigates employment ultrawideband (UWB) real-time location systems (RTLS) a factory environment proposes an augmentation technique to mitigate impairments that arise such complex scenario. A Bayesian...

Smart healthcare relies on artificial intelligence (AI) functions for learning and analysis of patient data. Since large diverse datasets training Machine Learning (ML) models can rarely be found in individual medical centers, classical centralized AI requires moving privacy-sensitive data from institutions to centers that process the fused information. Training thus higher communication resource/energy demands while violating privacy. This is considered today as a significant bottleneck...

Federated Learning (FL) systems orchestrate the cooperative training of a shared machine learning (ML) model across connected devices. Recently, decentralized FL architectures driven by consensus have been proposed to enable devices share and aggregate ML parameters via direct sidelink communications. The approach has advantage promoting federation among agents even in absence server, but may require an intensive use communication resources compared vanilla methods. This paper proposes...

This paper considers the problem of cooperative localization passive objects in a vehicular environment through fusion lidar point clouds collected at different moving vehicles and sent to road infrastructure. Object is then used improve position estimate according implicit positioning paradigm. At first, each vehicle uses deep neural network (a 3D object detector) process its cloud localize static objects. Then, set estimated bounding boxes infrastructure, which performs data association...

This paper considers the problem of cooperative lidar sensing in vehicular networks. We focus on task associating vehicle-generated measurements by lidars to enable a detection vulnerable road users. The considered are three-dimensional bounding boxes extracted from point cloud. Focusing centralized architecture which aggregates and processes all information, we design graph formulation association propose novel solution based Message Passing Neural Networks (MPNNs). method has advantage...

Accurate positioning is known to be a fundamental requirement for the deployment of Connected Automated Vehicles (CAVs). To meet this need, new emerging trend represented by cooperative methods where vehicles fuse information coming from navigation and imaging sensors via Vehicle-to-Everything (V2X) communications joint environmental perception. In line with trend, paper proposes novel data-driven sensing framework, termed Cooperative LiDAR Sensing Message Passing Neural Network (CLS-MPNN),...

This study addressed the problem of localization in an ultrawide-band (UWB) network, where positions both access points and tags needed to be estimated. We considered a fully wireless UWB system, comprising software hardware, featuring easy plug-and-play usability for consumer, primarily targeting sport leisure applications. Anchor self-localization was by two-way ranging, also embedding Gauss-Newton algorithm estimation compensation antenna delays, modified isolation forest working with...

Federated Learning (FL) techniques are emerging in the automotive context to support connected automated driving services. Yet, when applied vehicular use cases, conventional centralized FL policies show some drawbacks terms of latency and scalability. This paper focuses on decentralized solutions, which attempt overcome such limitations, by introducing a distributed computing architecture: vehicles exchange parameters shared Machine (ML) model via V2V links, without need central...

Bayesian Federated Learning (FL) offers a principled framework to account for the uncertainty caused by limitations in data available at nodes implementing collaborative training. In FL, exchange information about local posterior distributions over model parameters space. This paper focuses on FL implemented Device-to-Device (D2D) network via Decentralized Stochastic Gradient Langevin Dynamics (DSGLD), recently introduced gradient-based Markov Chain Monte Carlo (MCMC) method. Based...

The widespread proliferation of Internet Things (IoT) devices has pushed for the development novel transformer-based Anomaly Detection (AD) tools an accurate monitoring functionalities in industrial systems. Despite their outstanding performances, transformer models often rely on large Neural Networks (NNs) that are difficult to be executed by IoT due energy/computing constraints. This paper focuses introducing tiny AD make them viable solutions on-device AD. Starting from state-of-the-art...

In cooperative localization systems, the fusion of information from multiple sensing platforms is acknowledged to improve accuracy sensed targets. However, data association required perform inference non-trivial be solved. this context, we propose a graph formulation problem among unlabelled produced at different sensors in which run Message Passing Neural Network (MPNN). The proposed MPNN algorithm suits for centralized architectures where all are connected single processing unit. We...

In Industry 4.0, real-time location systems are emerging as a key technology to improve the efficiency of industrial processes, they allow track any assets or material movement and collect data on their usage. Ultra Wideband (UWB) offer unrivaled localization accuracy, but call for augmentation strategies in environments with complex propagation conditions such plants factories high density scattering objects obstructions. this paper, we focus Bayesian filtering techniques counterbalance...

The exchange of sensing information through Vehicle-to-Everything (V2X) communications enables the development cooperative systems for localization augmentation in connected automated vehicles. In V2X scenarios, integration measurements from multiple vehicles enhances environmental perception which is utmost importance enhanced safety services. this paper, we propose a Deep Neural Network (DNN)-assisted method that relies on centralized road infrastructure and network lidar sensors at...

Accurate positioning is known to be a fundamental requirement for the deployment of Connected Automated Vehicles (CAVs). To meet this need, new emerging trend represented by cooperative methods where vehicles fuse information coming from navigation and imaging sensors via Vehicle-to-Everything (V2X) communications joint environmental perception. In line with trend, paper proposes novel data-driven sensing framework, termed Cooperative LiDAR Sensing Message Passing Neural Network (CLS-MPNN),...

Federated Learning (FL) has emerged as a promising approach for privacy-preserving machine learning, particularly in sensitive domains such healthcare. In this context, the TRUSTroke project aims to leverage FL assist clinicians ischemic stroke prediction. This paper provides an overview of network infrastructure. The proposed architecture adopts client-server model with central Parameter Server (PS). We introduce Docker-based design client nodes, offering flexible solution implementing...

This paper addresses the problem of localization through Ultra Wide-Band (UWB) devices in case unavailability a permanent infrastructure, meaning that an adhoc UWB network has to be installed. Deploying system requires human intervention and calibration phases measure positions anchor nodes. In this paper, we propose iterative approach based on Gauss-Newton algorithm for addressing compensation antenna delay at each node anchors self-localization by two way ranging. The validation considers...

Poor accuracy of frequency references used in GSM base transceiver stations (BTS) can lead to dropped calls, slow handover between cells and even co-channel interference. In this paper, we report some results an experimental trial carried out the Omnitel-Vodafone test plant Milano (Italy). authors' knowledge, is first paper confirming with data that lack BTS synchronization affects performance, leading degradation quality service. The speech calls two handsets, undergoing BTSs synchronized...

In recent years, automotive systems have been integrating Federated Learning (FL) tools to provide enhanced driving functionalities, exploiting sensor data at connected vehicles cooperatively learn assistance information for safety and maneuvering systems. Conventional FL policies require a central coordinator, namely Parameter Server (PS), orchestrate the learning process which limits scalability robustness of training platform. Consensus-driven methods, on other hand, enable fully...

This paper focuses on highly precise localization and tracking of vehicles in race circuits, where centimeter-level accuracy is required for safety enabling complex maneuvering. Recently, the Annealed Stein Particle Filter (ASPF) has been proposed as a promising Bayesian tool tracking, showing its superior performances against conventional filtering methods, such Extended Kalman (EKF) (PF). Despite excellent performances, ASPF entails large computational complexity, making it unsuitable...