To support multiple AI-based applications, mobile systems need to collaboratively execute DNN architectures on heterogeneous AI accelerators. At the same time, the increasing DNN complexity and high degree of diversity in workloads on multichip module (MCM) accelerators are pushing AI processing off mobile nodes onto the edge. This has made computationally intensive, edge-based solutions the dominant approach for the deployment of modern neural networks. However, the rigid structure of fully-executed DNNs fails to align with the modular nature of MCM architectures, limiting their potential for efficient execution. In this paper, we introduce CLEAR, a novel optimization framework based on geometric programming that leverages both transformer-based and more canonical DNNs with early exits. CLEAR enables fast, coordinated decisionmaking across DNN design, workload distribution, and resource allocation, with the overarching goal of minimizing inference energy consumption. To our knowledge, this is the first work to integrate dynamic DNN optimization with decisions at both the communication infrastructure and hardware accelerator levels. We evaluate CLEAR using real-world wireless measurements and dynamic DNNs applied to computer vision inference tasks. Our results demonstrate that CLEAR achieves near-optimal performance and reduces energy consumption and resource usage by over 80% and 70%, respectively, compared to its benchmark.

The growing complexity of Cyber-Physical Systems (CPS) in industrial and manufacturing environments calls for more sophisticated methods to represent heterogeneous assets and processes. In response, hierarchical Digital Twins (DTs)–virtual representations of physical, taxonomy-based processes–offer transparent, layered modeling of diverse data sources. This layered structure fuels renewed interest in intelligent engines capable of extracting meaningful insights and mapping them within the stratified DT ecosystem. While current Intelligent Digital Twin (I-DT) engines based on Deep Learning are computationally demanding, lightweight alternatives like Reservoir Computing (RC) offer efficient solutions with low training costs and fast inference for modeling causal dynamics. This inherent trade-off between performance and practicality underscores the limitations of evaluating I-DTs on accuracy alone. To address this gap, this work introduces a novel metric, Fidelity , designed to provide a comprehensive evaluation. Unlike traditional approaches, Fidelity also accounts for maintainability and deployability, especially in contexts involving time-varying and hierarchical data dynamics. Extensive experiments on two multimodal datasets demonstrate the competitiveness of our RC-based engine and highlight the value of introducing Fidelity for effectively profiling I-DTs. Specifically, our RC-based engine, identified as optimal through a higher Fidelity score, consumes an order of magnitude less energy and achieves up to 39 % higher accuracy (about 10 % increase on average) compared to both canonical and other RC-based alternatives.

Computer Vision (CV) tasks are among the most pivotal, yet challenging, operations for Uncrewed Aerial Vehicles (UAVs), especially in mission-critical applications. They require processing complex image data through Deep Neural Networks (DNNs), which demand computational resources far beyond UAVs’ capacity. To address this limitation, Split DNNs offer a promising solution by partitioning the model into: (i) a lightweightHead, deployed on the UAV for rapid, albeit less precise, initial image representations, and (ii) a more complexTail, executed at the network edge for refined, higher-accuracy results. However, this solution necessitates transmitting large tensor data from the UAV to the edge server, leading to significant bandwidth consumption. We tackle this challenge by introducing a goal-oriented framework named Compressed Tensor-based DNN Split (CoTeD). Our framework integrates an application- and system-aware optimization model that orchestrates computing and transmission resources in real time. At the UAV, CoTeD dynamically selects relevant tensor information and optimally trades-off between DNN detection quality and bandwidth consumption, guided by application requirements and system operational conditions. At the edge server, CoTeD reconstructs the tensor, enabling efficient inference by the Tail model. This approach effectively balances bandwidth usage with quality of the CV task output. Experimental results, obtained through our hardware-software testbed and using datasets with different sizes and characteristics, show that CoTeD can reduce data transmission over the radio link by up to 90% without noticeable loss in object detection quality and inference latency by up to 70% compared to local DNN deployment onboard the UAV. Also, CoTeD yields an inference request success rate of at least 90%, with an increase of 20%-80% compared to direct DNN splitting, static JPEG compression, and DNN model quantization.

The remarkable success of attention-based models in real-world applications has sparked a crucial question for Reservoir Computing (RC): Can its inherent computational efficiency compete with the high-performance, yet energy-intensive, novel deep learning architectures? Can Deep and modular RC neural networks address state-of-the-art challenges in Computer Vision and Natural Language Processing? In the attempt to consolidate RC capabilities towards more complex tasks, this paper delves into the exploration of a comprehensive RC’s offline-online cycle cost analysis. Our investigation highlights hyperparameters (HPs) optimization as a major bottleneck in RC deployment, particularly for those exploring RC capabilities and those who want to maintain user-level knowledge of the solution. To address this, we introduce an adaptive ϵ-Greedy based search exploration mechanism, significantly streamlining the off-line optimization process while maintaining high accuracy. Furthermore, we enhance existing RC frameworks to support online transfer learning and inference, enabling seamless, fast, and energy-efficient adaptation to real-world environments. By analyzing the impact of optimized HPs on performance, we aim to demonstrate the viability of RC as a powerful and efficient alternative for many practical applications, including those on devices with limited resources. Experimental results proved that our solution is able to reduce the time required for offline HPs optimization by 70%, enabling energy savings of up to 88%. Moreover, in the online scenario, it guarantees similar performance in terms of accuracy while reducing memory usage by 66%.

Split computing (≠ split learning) is a promising approach to deep learning models for resource-constrained edge computing systems, where weak sensor (mobile) devices are wirelessly connected to stronger edge servers through channels with limited communication capacity. State-of-the-art work on split computing presents methods for single tasks such as image classification, object detection, or semantic segmentation. The application of existing methods to multi-task problems degrades model accuracy and/or significantly increase runtime latency. In this study, we propose Ladon, the first multi-task-head supervised compression model for multi-task split computing.11Code and models are available at https://github.com/yoshitomo-matsubara/ladon-multi-task-sc2 Experimental results show that the multi-task supervised compression model either outperformed or rivaled strong lightweight baseline models in terms of predictive performance for ILSVRC 2012, COCO 2017, and PASCAL VOC 2012 datasets while learning compressed representations at its early layers. Furthermore, our models reduced end-to-end latency (by up to 95.4%) and energy consumption of mobile devices (by up to 88.2%) in multi-task split computing scenarios.

Real-world applications requiring real-time responsiveness frequently rely on energy-intensive and compute-heavy neural network algorithms. Strategies include deploying distributed and optimized Deep Neural Networks on mobile devices, which can lead to considerable energy consumption and degraded performance, or offloading larger models to edge servers, which requires low-latency wireless channels. Here we present Furcifer, a novel middleware that autonomously adjusts the computing strategy (i.e., local computing, edge computing, or split computing) based on context conditions. Utilizing container-based services and low-complexity predictors that generalize across environments, Furcifer supports supervised compression as a viable alternative to pure local or remote processing in real-time environments. An extensive set of experiments coversdiverse scenarios, including both stable and highly dynamic channel environments with unpredictable changes in connection quality and load. In moderate-varying scenarios, Furcifer demonstrates significant benefits: achieving a 2x reduction in energy consumption, a 30% higher mean Average Precision score compared to local computing, and a three-fold FPS increase over static offloading. In highly dynamic environments with unreliable connectivity and rapid increases in concurrent clients, Furcifer’s predictive capabilities preserves up to 30% energy, achieving a 16% higher accuracy rate, and completing 80% more frame inferences compared to pure local computing and approaches without trend forecasting, respectively. • Adaptive Split Computing: an efficient strategy for real-time vision applications. • A Middleware for automated, dynamic, and resilient flexible computing. • A Low-Complexity Manager for efficient power, higher FPS rate, and enhanced accuracy.

Modern real-time applications widely embed compute intense neural algorithms at their core. Current solutions to support such algorithms either deploy highly-optimized Deep Neural Networks at mobile devices or offload the execution of possibly larger higher-performance neural models to edge servers. While the former solution typically maps to higher energy consumption and lower performance, the latter necessitates the low-latency wireless transfer of high volumes of data. Time-varying variables describing the state of these systems, such as connection quality and system load, determine the optimality of the different computing configurations in terms of energy consumption, task performance, and latency. Herein, we propose Furcifer, a framework capable of dynamically adapting the cloud continuum computing configuration in response to the perceived state of the system. Our container-based approach incorporates low-complexity predictors that generalize well across operating environments. In addition, we develop a highly optimized split Deep Neural Network model, which achieves in-model supervised compression and enhances task offloading. Experimental results for object detection across diverse conditions, environments, and wireless technologies, show Furcifer's remarkable outcomes, including a 2x energy reduction, 30% higher mean Average Precision score than pure local computing, and a notable three-fold increase in frame per second rate compared to static offloading.

Federated Learning (FL) represents the de-facto standard paradigm for enabling distributed learning over multiple clients in real-world scenarios. Despite the great strides reached in terms of accuracy and privacy awareness, the real adoption of FL in real-world scenarios, in particular in industrial deployment environments, is still an open thread. This is mainly due to privacy constraints and to the additional complexity stemming from the set of hyperparameters to tune when employing AI techniques on bandwidth-, computing-, and energy-constrained nodes. Motivated by these issues, we focus on scenarios where participating clients are characterised by highly heterogeneous computing capabilities and energy budgets proposing EneA-FL, an innovative scheme for serverless smart energy management. This novel approach dynamically adapts to optimize the training process while fostering seamless interaction between Internet of Things (IoT) devices and edge nodes. In particular, the proposed middleware provides a containerised software module that efficiently manages the interaction of each worker node with the central aggregator. By monitoring local energy budget, computational capabilities, and target accuracy, EneA-FL intelligently takes informed decisions about the inclusion of specific nodes in the subsequent training rounds, effectively balancing the tripartite trade-off between energy consumption, training time, and final accuracy. Finally, in a series of extensive experiments across diverse scenarios, our solution demonstrates impressive results, achieving between 30% and 60% lower energy consumption against popular client selection approaches available in the literature while being up to 3.5 times more efficient than standard FL solutions.

The increasing availability of data generated by Internet of Things (IoT) and Industrial Internet of Things (IIoT) devices, as well as privacy and law regulations, have significantly boosted the interest in collaborative machine learning (ML) approaches. In this direction, we claim federated learning (FL) as a promising ML paradigm where participants collaboratively train a global model without outsourcing on-premises data. However, setting up and using FL can be extremely costly and time-consuming. To effectively promote the adoption of FL in real-world scenarios, while limiting the overhead and knowledge of the underlying technology, service providers should offer federated learning as a service (FLaaS). One of the major concerns while designing an architecture that provides FLaaS is achieving trustworthiness among involved typically unknown participants. This article presents a blockchain-based architecture that achieves Trustworthy federated learning as a service (TruFLaaS). Our solution provides trustworthiness among 3rd-party organizations by leveraging blockchain, smart contracts, and a decentralized oracle network. Specifically, during each FL round, the service provider supplies a sample, without overlapping, of its validation set to validate all partial models submitted by clients. By doing so, poor models, which tend to degrade performance or introduce malicious backdoors, are identified and discarded. Due to the transparency of the blockchain, not changing the validation set would enable participants to forge a malicious partial model that passes the validation phase. We evaluate our approach over two well-known IIoT datasets: the reported experimental results show that TruFLaaS outperforms the state-of-the-art literature solutions in the field.

Federated Learning (FL) has emerged as a value added proposition for use in edge-based infrastructures, distributing the training process among collaborative workers without disclosing raw (user) data. In this context, we argue that, differently from what already present in most current literature, energy consumption of nodes (either workers or the ensembler node) is a central element to consider in FL, e.g., to have a more sustainable FL node selection strategy. To this end, a complete and detailed report about energy consumption at each FL round is required to allow for innovative and greener resource management approaches, taking into account residual energy and learning completion time of participating FL nodes. Filling this gap, we present the design of a novel distributed framework capable of collecting accurate (worker) energy expenditure and learning-centric metrics at each FL round. The frame-work comprises state-of-the-art technological building blocks, purposely integrated to enable advanced and energy-aware FL process orchestration capabilities. To validate the approach, we rely on a heterogeneous experimental testbed, and conduct a distributed learning process employing a realistic dataset. The preliminary evaluation results reported in this paper highlight the potential advantage in terms of overall energy consumption reduction and the suitability of an adaptive learning framework capable of autonomous evaluations of the most proper trade-off to apply between accuracy and energy expenditure.

The increasing pace of sensing and communication technology rollout is paving the way for concrete deployments of smart city applications, enabling a data-driven modeling of processes and the environment. In particular, the Urban Facility Management (UFM) process is growing in importance, recognized to have a direct impact on the sustainability and the development of our cities. In [1] we presented a system's view of a Digital Twin solution for the UFM process. The solution relies on (near)real-time data to quantify the activity index in an area of interest, used as a basis for planning decisions. In this study, we focus on the predictive subsystem, tasked with computing near-to-mid term predictions of the activity index, equipping UFM operators with a flexible decision-support system. Without loss of generality, we present an analysis of the vehicular traffic component, part of the activity index, assessing the accuracy of different predictive schemes, discussing some operational implications.

The application of Internet of Things (IoT) within industrial environments is fostering the adoption of the digital twin (DT) approach, applied at the edge of the network to handle heterogeneity stemming from siloed application management solutions and from protocols originated by different manufacturing tools and enterprise services. In this challenging context, network heterogeneity also represents a critical element that can significantly limit the design and deployment of DT-oriented applications. This article proposes the Application-driven digital twin networking middleware with the twofold objective of: 1) Simplifying the interaction among heterogeneous devices by allowing DTs to exploit IP-based protocols instead of specialized industrial ones and to enhance packet content expressiveness, by enriching data via well-defined standards. 2) Dynamically managing network resources in edge industrial environments, applying software defined networking to exploit the communication mechanisms most suitable to application requirements, ranging from native IP to more articulated based on packet content.

The smart home concept can significantly benefit from predictive models that take proactive management operations on home actuators, based on users' behavior evaluation. In this paper, we use a small-scale physical model, the ScaledHome-2 testbed, to experiment with the evolution of measurements in a suburban home under different environmental scenarios. We start from the observation that, for a home to become smart, in addition to IoT sensors and actuators, we also need a predictive model of how actions taken by inhabitants and home actuators affect the internal environment of the home, reflected in the sensor readings. In this paper, we propose a technique to create such a predictive model through machine learning in various simulated weather scenarios. This paper also contributes to the literature in the field by quantitatively comparing several machine learning algorithms (K-nearest neighbor, regression trees, Support Vector Machine regression, and Long Short Term Memory deep neural networks) in their ability to create accurate and generalizable predictive models for smart homes.