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.

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.

Machine learning (ML) can be often distributed, owing to the need to harness more resources and/or to preserve privacy. Accordingly, distributed learning has received significant attention from the literature; however, most works focus on the expected learning quality (e.g., loss) attained and do not consider the distribution thereof. It follows that ML models are not dependable, and may fall short of the required performance in many real-world cases. In this work, we tackle this challenge and propose DepL, a framework attaining dependable learning orchestration. DepL efficiently makes joint, near-optimal decisions concerning (i) which data to use for learning, (ii) the ML models to use – chosen within a set of full-size models and compressed versions thereof – and when to switch from one model to another, and (iii) the clusters of physical nodes to use for the learning. DepL improves over previous works by guaranteeing that the learning quality target (e.g., a minimum loss) is achieved with a target probability, while minimizing the learning (e.g., energy) cost. DepL has provably low polynomial computational complexity and a constant competitive ratio. Further, experimental results using the CIFAR-10 and GTSRB datasets show that it consistently matches the optimum and outperforms state-of-theart approaches (30% faster learning and 40–80% lower cost).

Distributed sensor fusion is a key component of a broad spectrum of applications, such as autonomous systems, where the ability of jointly process multi-sensor data at the edge boosts the range of operating conditions and overall task performance. However, existing distributed sensor fusion approaches encounter limitations in achieving efficient transmission and computation, primarily due to sensor data redundancy, unreliable sensor data transmission, and inflexible sensor fusion methods. In this paper, we propose HERACLES, a distributed sensor fusion framework that connects multi-branched dynamic neural network architectures, which we extend to include branches of different complexity, to (i) a computing methodology that distributes portions of the multi-branched neural network across mobile devices and edge servers, enabling flexible semantic feature extraction and sensor fusion; (ii) a hierarchical modulation-based transmission strategy, where multi-modal semantic features are allocated to different modulation layers to provide varying levels of error protection, and (iii) an infrastructure-level logic that controls the matching between semantic features and modulation layers, and the complexity of the neural model itself to meet an accuracy target while minimizing latency and energy consumption. As a result, HERACLES deeply connects computing, communications and resource allocations in a semantic and context-aware fashion. We evaluate HERACLES using real-world datasets and demonstrate that it can reduce the total delay and energy consumption by 20.39%–89.41% and 4.86%–88.17% (resp.), while maintaining near-optimal inference accuracy. The evaluation code is available at https://github.com/qlt315/HERACLES.

Split Computing has emerged as a promising paradigm for deploying Deep Neural Networks in Edge and Inter-net of Things systems, enabling inference tasks to be distributed between resource-constrained edge devices and cloud servers. This approach is particularly attractive for autonomous systems, where security and reliability may be critical. However, interme-diate feature maps transmitted between devices are vulnerable to corruption, which may result from intentional adversarial attacks or unintentional hardware faults. Distinguishing whether corruption originates from an external adversary or an inherent system fault is crucial for implementing appropriate counter-measures-reinforcing security mechanisms against attacks or improving system reliability to mitigate the effects of hardware-related faults. To the best of our knowledge, this work is the first to propose a machine learning-based classification mechanism capable of differentiating adversarial attacks from hardware defects in Split Computing systems. The proposed approach analyzes the intermediate feature maps transmitted from the edge device to the server, classifying the source of corruption to guide appropriate responses. Experimental results demonstrate that one of the proposed classifiers can distinguish between intentional and unintentional feature map corruptions with an accuracy of 93.91 %.

The fusion of multi-modal information, such as images and LiDAR scans, is instrumental to maximize the performance of many computer vision tasks in next generation systems and applications. However, supporting fusion necessitates considerable effort, challenging the availability of computing and communication resources in edge systems. This work addresses this challenge by maximizing resource efficiency in systems where mobile devices collect multi-modal sensor data and use dynamic multi-branched DNN models to adapt inference to the operating context. To tune the overall system response to the context (e.g., weather conditions), we propose a dual-scale control approach: centralized orchestration of spectrum resources, and distributed individual device-level control of the execution path of the dynamic DNN fusion models. The control agents are driven by a novel context-aware decision-making method combined with game theory, named Context-Aware Network Slicing Auction (CANSA), which optimizes DNN inference performance, network slicing, and energy consumption. The decision-making performs such optimization by: (i) selecting data and features that best fit the current context; (ii) deciding on the appropriate DNN model complexity, including the use of multi-modal sensor fusion techniques for better data integration, and (iii) deploying these models on the most appropriate nodes (local nodes or edge servers). Results, obtained using real-world multi-modal data, show that CANSA surpasses conventional allocation methods by up to 52.3% in terms of inference task success rate.

Connected Autonomous Vehicles (CAVs) are equipped with an array of sensors generating substantial data streams whose real-time analysis often exceed the onboard computational capabilities. While offloading these computational tasks to edge servers is an established solution, such an approach is increasingly challenging as multiple sensor streams need to be fused and analyzed, and thus transferred over capacity-limited and volatile wireless channels. To address this challenge, in this paper we propose a "split computing" framework. Compared to existing solutions that focus on individual sensor streams (and most commonly cameras), our framework is designed for sensor fusion neural models, and specifically camera and LiDAR fusion for semantic segmentation. Our proposed framework optimizes data transfer by eliminating pooling indices in favor of sending LiDAR point cloud indices only to the final network block. We remark how sensor fusion is instrumental to guarantee robust operations in a broad range of conditions. By compressing the data to be transported over the channel, our approach reduces offloading latency, better utilizes CAV computational resources compared to full offloading schemes and decreases channel load in congested urban network scenarios. Our experimental results demonstrate up to 62.74% improvement in task execution latency for sensor fusion models, with at most 6.61% performance trade-off due to compression.

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.

In response to the escalating demand for high-performance applications in Connected Autonomous Vehicles (CAVs), there has been an increased effort to accelerate Computer Vision (CV) tasks. Innovations in the field suggest that partitioning Deep Learning (D L) models, which are pivotal for CV tasks, can effectively utilize the computational strengths of both CAVs and proximal edge servers. This is achieved by splitting the DL models into two parts, where execution begins on the CAV, then the intermediate results are offloaded to complete processing on the edge server. Moreover, the precision and response time of these applications vary significantly, necessitating the adoption of early exit strategies. These strategies involve modifying DL models with auxiliary branches that allow for early termination at various stages, enabling the model to exit processing early when application requirements are met. To navigate the trade-offs between accuracy and latency, it is imperative to choose the model's optimal split and exit point that align with the current environmental conditions, such as server data throughput and computational load. To achieve this, we employ a state-of-the-art method that utilizes the Deep Deterministic Policy Gradient (DDPG) algorithm, augmented by Long Short-Term Memory networks. Additionally, to refine the training of the DDPG's actor network, Imitation Learning (IL) is integrated. IL employs a heuristic exhaustive search to retrospectively identify the most effective model partition and exit point.

Mobile computing faces challenges due to the resource constraints of mobile devices, such as limited computing power, energy, and connectivity. These limitations hinder the use of high-complexity classifiers and wireless transmissions. To address this issue, we propose a novel collaboration paradigm between mobile devices and edge servers, where the edge server assists the mobile devices by dynamically retraining a low-complexity classifier to adapt to temporal changes in data distribution. We propose a novel approach called drift control protocol (DCP) which is inspired by TCP congestion control mechanism. DCP aims to strike a balance between low-complexity classifier retraining frequency and communication costs with the edge server. It adjusts the update rate of the classifier on the mobile device based on distribution drift characteristics and controls the number of input samples sent to the edge server to improve accuracy. We evaluate and study different versions of DCP using synthetic and real datasets We demonstrate that DCP keeps the error bound, while reducing the burden of the communication cost by 90% for the mobile nodes, which makes our proposal suitable for online domain adaptation.

Lightweight autonomous unmanned aerial vehicles (UAV) are emerging as a central component of a broad range of applications. However, autonomous navigation necessitates the implementation of perception algorithms, often deep neural networks (DNN), that process the input of sensor observations, such as that from cameras and LiDARs, for control logic. The complexity of such algorithms clashes with the severe constraints of these devices in terms of computing power, energy, memory, and execution time. In this paper, we propose NaviSplit, the first instance of a lightweight navigation framework embedding a distributed and dynamic multi-branched neural model. At its core is a DNN split at a compression point, resulting in two model parts: (1) the head model, that is executed at the vehicle, which partially processes and compacts perception from sensors; and (2) the tail model, that is executed at an interconnected compute-capable device, which processes the remainder of the compacted perception and infers navigation commands. Different from prior work, the NaviSplit framework includes a neural gate that dynamically selects a specific head model to minimize channel usage while efficiently supporting the navigation network. In our implementation, the perception model extracts a 2D depth map from a monocular RGB image captured by the drone using the robust simulator Microsoft AirSim. Our results demonstrate that the NaviSplit depth model achieves an extraction accuracy of 72.81 % while transmitting an extremely small amount of data (1.218 KB) to the edge server. When using the neural gate, as utilized by NaviSplit, we obtain a slightly higher navigation accuracy as compared to a larger static network by 0.3% while significantly reducing the data rate by 95%. To the best of our knowledge, this is the first exemplar of dynamic multi-branched model based on split DNNs for autonomous navigation.

Theexisting work on the distributed training of machine learning (ML) models has consistently overlooked the distribution of the achieved learning quality, focusing instead on its average value. This leads to a poor dependability of the resulting ML models, whose performance may be much worse than expected. We fill this gap by proposing DepL, a framework for dependable learning orchestration, able to make high-quality, efficient decisions on (i) the data to leverage for learning, (ii) the models to use and when to switch among them, and (iii) the clusters of nodes, and the resources thereof, to exploit. For concreteness, we consider as possible available models a full DNN and its compressed versions. Unlike previous studies, DepL guarantees that a target learning quality is reached with a target probability, while keeping the training cost at a minimum. We prove that DepL has constant competitive ratio and polynomial complexity, and show that it outperforms the state-of-the-art by over 27% and closely matches the optimum.

Recent advancements in 6G technologies have revolutionized vehicle-to-everything communications, accelerating the evolution of autonomous driving (AD) from a conceptual vision to an operational reality. In this paper, we propose an agile architecture integrated with an offloading framework specifically designed to support AD tasks by harnessing the potential of 6G, while adhering to ISO standards for Automotive Safety Integrity Level (ASIL). Unlike existing offloading approaches that rely on edge/cloud, the proposed ASIL-Centric Offloading Framework for Autonomous Driving (ACOFAD) introduces roadside units as additional remote computing nodes and incorporates unmanned aerial vehicles as communication relays to extend coverage in hard-to-reach or underserved areas. Offloading decisions are dynamically made based on node availability, link reliability, latency constraints, and task criticality. Simulation results demonstrate that ACOFAD achieves very low latency, evaluated to 5 ms and 0.5 s for 20 MB and 2GB payloads, respectively, enabling efficient offload of both safety-critical and non-critical tasks.

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.

Next-generation mobile systems will support multiple AI-based applications, each leveraging heterogeneous sensors and data sources through deep neural network (DNN) architectures collaboratively executed within the network. In this context, to minimize the cost of the AI inference task subject to requirements on latency, quality, and – crucially –reliabilityof the inference process, it is vital to optimize (i) the set of sensors/data sources and (ii) the DNN architecture, (iii) the network nodes executing sections of the DNN, and (iv) the resources to use. To achieve these goals, we leverage dynamic gated neural networks with branches, and propose a novel algorithmic strategy called Quantile-constrained Inference (QIC), based upon quantile-Constrained policy optimization. QIC makes joint, high-quality, swift decisions on all the above aspects of the system, with the aim to minimize inference energy cost. We remark that this is the first contribution connecting gated dynamic DNNs with infrastructure-level decision making. We evaluate QIC using a dynamic gated DNN with stems and branches for optimal sensor fusion and inference, trained on the RADIATE dataset offering Radar, LiDAR, and Camera data, and real-world wireless measurements. Our results confirm that QIC closely matches the optimum and outperforms existing approaches in reducing energy consumption (compute, communication, and total) and application requirements failure by over 70%.

An increasing number of mobile applications leverage deep neural networks (DNN) as an essential component to adapt to the operational context at hand and provide users with an enhanced experience. It is thus of paramount importance that network systems support the execution of DNN inference tasks in an efficient and sustainable way. Matching the diverse resources available at the mobile-edge-cloud network tiers with the applications requirements and the complexity of their, while minimizing energy consumption, is however challenging. A possible approach to the problem consists in exploiting the emerging concept of dynamic DNNs, characterized by multi-branched architectures with early exits enabling sample-based adaptation of the model depth. We leverage this concept and address the problem of deploying portions of DNNs with early exits across the mobile-edge-cloud system and allocating therein the necessary network, computing, and memory resources. We do so by developing a 3-stage graph-modeling method that allows us to represent the characteristics of the system and the applications as well as the possible options for splitting the DNN over the multi-tier network nodes. Our solution, called Feasible Inference Graph (FIN), can determine the DNN split, deployment, and resource allocation that minimizes the inference energy consumption while satisfying the nodes’ constraints and the requirements of multiple, co-existing applications. FIN closely matches the optimum and leads to over 89% energy savings with respect to state-of-the-art alternatives.

Early Exiting (EE) is an emerging computing paradigm where Deep Neural Networks (DNNs) are equipped with earlier classifiers, enabling trading-off accuracy with inference latency. EE can be effectively combined with edge computing, a paradigm that allows mobile nodes to offload complex tasks, such as the execution of DNNs, to servers at the edge of the network, thus reducing computing times and energy consumption at the mobile devices. The integration of such technologies is particularly attractive for the support of applications for connected and automated driving. In this paper, we consider a system that jointly leverages the benefits of EE and edge computing, and we model their complex interactions by means of a Markov Decision Process (MDP). We then formulate an optimization problem to select the inference strategy that maximizes the average task accuracy. Importantly, such an optimization problem has low complexity, as the optimal policy can be derived by mapping the MDP into a linear program. Our numerical results focus on a use case centered on automated vehicles connected with an edge server under varying channel and network conditions, and show that our solution achieves up to 11% higher accuracy compared to the optimal policy with no EE.

The increasing pervasiveness of intelligent mobile applications requires to exploit the full range of resources offered by the mobile-edge-cloud network for the execution of inference tasks. However, due to the heterogeneity of such multi-tiered networks, it is essential to make the applications’ demand amenable to the available resources while minimizing energy consumption. Modern dynamic deep neural networks (DNN) achieve this goal by designing multi-branched architectures where early exits enable sample-based adaptation of the model depth. In this paper, we tackle the problem of allocating sections of DNNs with early exits to the nodes of the mobile-edge-cloud system. By envisioning a 3-stage graph-modeling approach, we represent the possible options for splitting the DNN and deploying the DNN blocks on the multi-tiered network, embedding both the system constraints and the application requirements in a convenient and efficient way. Our framework – named Feasible Inference Graph (FIN) – can identify the solution that minimizes the overall inference energy consumption while enabling distributed inference over the multi-tiered network with the target quality and latency. Our results, obtained for DNNs with different levels of complexity, show that FIN matches the optimum and yields over 65% energy savings relative to a state-of-the-art technique for cost minimization.

Artificial intelligence is becoming increasingly popular for IoT applications in safety-critical fields (e.g., autonomous systems and biomedical, robots). Unfortunately, the inference’s workload process alone increases as the model size grows. To meet the computational power limitations of mobile devices running IoT applications, modern services sometimes resort to the Split Computing paradigm. Split Computing divides the inference process of a Neural Network into Head and Tail for their execution in a mobile device and a server, respectively, which also allows the reduction of the overall IoT device’s computational cost. Nonetheless, Split Computing can be used in safety-critical fields where reliability is crucial, especially when mobile devices have computational and cost restrictions. This paper introduces hardening techniques acting on the software to mitigate the effects of hardware faults on Split Computing models. The proposed hardening techniques consist of i) a bounded activation function whose thresholds are refined by training, and ii) a per-channel bounding of the bottleneck quantization of the split points. To quantitatively assess their effectiveness, we resorted to two different split configurations of a model for image classification. In addition, we considered a Split Computing model for object detection. Our findings indicate that the proposed approaches effectively reduces fault effects by $\mathbf{3. 5 \%}$ for image classifiers and $5.73 \%$ for object detectors when compared with other hardening approaches for general DNNs.

The deployment of Machine Learning (ML) applications extensively leverages Matrix Multiplication (MM) operations on modern and advanced accelerators, like Graphic Processing Units (GPUs), which employ Tensor Core Units (TCUs) to optimize MM’s execution efficiently. However, reliability concerns arise in devices with cutting-edge semiconductor technologies (7 nm or less), as faults can compromise some structures (e.g., TCUs) during their operation. In safety-critical applications, this can lead to wrong DNN outcomes and cause unpredictable and unacceptable actions. Thus, the impact evaluation of such faults is crucial to ensure that TCUs and GPUs meet the safety standard requirements (e.g., ISO26262). Currently, the reliability assessment of complex applications concerning hardware faults involves fault injection (FI) campaigns. Unfortunately, low-level FI campaigns might be computationally prohibitive for GPUs when these execute massive applications like DNNs. In this work, we propose an error modeling approach to accurately describe corruptions from permanent faults on TCUs, during the operation of MMs. This approach enables realistic reliability evaluations of computationally expensive MM-based workloads, resulting in a huge acceleration (up to 225X) compared with hardware-level FIs. Our experimental results show a very good accuracy (up to $93 \%$ correlation between our error modeling approach and FI campaigns conducted on TCUs).

Mobile robots suffer from inherent limitations due to the tradeoff in the amount of energy consumed by their on-board processing components, and the need to increase their operational time. On the communication side, the volatility of communication links severely hinders the ability of a mobile device to rely on computation offloading. The challenge addressed by this paper is the development of a methodology and framework to effectively migrate the location of a service from a system to another, minimizing downtime and striving to reduce any side-effects that may be perceived by the system. Solving this challenge will pave the way for more effective computation offloading solutions that can cope with the unpredictability of the edge systems. Four different approaches are compared, analyzing their performance via an empirical approach. The insights gathered from data allow the identification of the most promising solution to address the aforementioned challenge.

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.

Mobile systems will have to support multiple AI-based applications, each leveraging heterogeneous data sources through DNN architectures collaboratively executed within the network. To minimize the cost of the AI inference task subject to requirements on latency, quality, and - crucially - reliability of the inference process, it is vital to optimize (i) the set of sensors/data sources and (ii) the DNN architecture, (iii) the network nodes executing sections of the DNN, and (iv) the resources to use. To this end, we leverage dynamic gated neural networks with branches, and propose a novel algorithmic strategy called Quantile-constrained Inference (QIC), based upon quantile-Constrained policy optimization. QIC makes joint, high-quality, swift decisions on all the above aspects of the system, with the aim to minimize inference energy cost. We remark that this is the first contribution connecting gated dynamic DNNs with infrastructure-level decision making. We evaluate QIC using a dynamic gated DNN with stems and branches for optimal sensor fusion and inference, trained on the RADIATE dataset offering Radar, LiDAR, and Camera data, and real-world wireless measurements. Our results confirm that QIC matches the optimum and outperforms its alternatives by over 80%.

The metaverse is a virtual space that combines physical and digital elements, creating immersive and connected digital worlds. For autonomous mobility, it enables new possibilities with edge computing and digital twins (DTs) that offer virtual prototyping, prediction, and more. DTs can be created with 3D scene reconstruction methods that capture the real world's geometry, appearance, and dynamics. However, sending data for real-time DT updates in the metaverse, such as camera images and videos from connected autonomous vehicles (CAVs) to edge servers, can increase network congestion, costs, and latency, affecting metaverse services. Herein, a new method is proposed based on distributed radiance fields (RFs), multi-access edge computing (MEC) network for video compression and metaverse DT updates. RF-based encoder and decoder are used to create and restore representations of camera images. The method is evaluated on a dataset of camera images from the CARLA simulator. Data savings of up to 80% were achieved for H.264 I-frame - P-frame pairs by using RFs instead of I-frames, while maintaining high peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) qualitative metrics for the reconstructed images. Possible uses and challenges for the metaverse and autonomous mobility are also discussed.

Recent advances in Internet-of-things (IoT) and 5G infrastructures promote new computational paradigms such as Split Computing (SC) for deploying Deep Neural Networks (DNNs) on mobile applications. In SC, DNNs are partitioned into head and tail sub-models that are executed on the mobile device and cloud/edge servers, respectively. Modern SC models resort to head compression techniques to balance energy consumption, transmission data, and model size while preserving the outstanding accuracy of large state-of-the-art DNNs. These features make SC DNNs suitable for mobile applications, including safety-critical systems (e.g., self-driving vehicles, autonomous robots, and healthcare equipment), where reliability is a paramount factor mandated by strict safety standards. Despite there are many studies available about the reliability of DNNs, the SC models are still unexplored, especially when hardware faults threaten the operation of a mobile device. In this work, we present for the first time $i)$ an application-level fault injection strategy for modeling hardware faults on mobile GPUs executing SC DNNs and ii) an evaluation of the resilience of supervised compression methods utilized by SC systems. The preliminary results gathered on some representative benchmark networks and configurations show the feasibility and effectiveness of the approach. They also demonstrate that aggressive compression strategies lead to high accuracy degradation $(\approx$ 40%), increasing the overall vulnerability of the DNN and the system.

The proliferation of airborne drones, while instrumental to a broad range of applications, has led to an increased number of regulatory non-compliance incidents. The ubiquitous unmanned aerial vehicles (UAVs) are posing security risks, since they have started to be used for cybercrimes. Effective detection of illicit drones in restricted areas is paramount. Evolved drones are more and more sophisticated, and sometimes they do not emit RF-based signals, making inapplicable RF-based detection solutions. Different from existing work, this paper introduces a neural sensor fusion framework for drone detection based on both audio and video data to accurately identify drones and differentiate them from similar objects at long distances. Our design adopts a late fusion approach using the Weighted Average and Random Forest algorithm on the visual and auditory classification pipeline. Specifically, we process infrared data using a You Only Look Once (YOLO) v5 model due to its balance between inference time and accuracy. For the audio stream, we evaluate Long Short-Term Memory (LSTM) and Convolutional Recurrent Neural Network (CRNN) models and demonstrate the superiority of the CRNN model through Mel-Frequency Cepstral Coefficients (MFCC) features. To demonstrate the robustness of our audio-visual fusion approach, we validate it in extensive scenarios, with impaired audio/video data. Our results demonstrate that multimodal fusion significantly improves drone detection, outperforming traditional single-modality systems in complex environments. Additionally, our system provides predictions rapidly, in just 0.382 seconds, making it well-suited for real-time applications.

Model compression is a fundamental tool to execute machine learning (ML) tasks on the diverse set of devices populating current-and next-generation networks, thereby exploiting their resources and data. At the same time, how much and when to compress ML models are very complex decisions, as they have to jointly account for such aspects as the model being used, the resources (e.g., computational) and local datasets available at each node, as well as network latencies. In this work, we address the multi-dimensional problem of adapting the model compression, data selection, and node allocation decisions to each other: our objective is to perform the DNN training at the minimum energy cost, subject to learning quality and time constraints. To this end, we propose an algorithmic framework called PACT, combining a time-expanded graph representation of the training process, a dynamic programming solution strategy, and a data-driven approach to the estimation of the loss evolution. We prove that PACT’s complexity is polynomial, and its decisions can get arbitrarily close to the optimum. Through our numerical evaluation, we further show how PACT can consistently outperform state-of-the-art alternatives and closely matches the optimal energy consumption.

The autonomous operations of unmanned aerial vehicles (UAV) necessitate the real-time analysis of information-rich signals, such as camera and LiDAR feeds, where the analysis algorithms often take the form of extremely complex deep neural networks (DNN). The continuous execution of such models onboard the UAV imposes a considerable resource consumption (e.g., energy), while offloading the execution of the models to edge servers requires the transmission of the input signals over capacity constrained, time-varying, wireless channels. In this paper, we propose an innovative approach – CADET – to control where sensor signals are processed in the system. In addition to traditional features and measures, such as channel state, energy consumption and channel usage, CADET makes dynamic task routing decisions – local computing vs edge computing – based on the state of the flight controller. The proposed methodology is based on Markov jump switched linear systems, where an embedded filter predicts and controls the state of the joint motion/computing dynamics.

Modern IEEE 802.11 (Wi-Fi) networks extensively rely on multiple-input multiple-output (MIMO) to significantly improve throughput. To correctly beamform MIMO transmissions, the access point needs to frequently acquire a beamforming matrix (BM) from each connected station. However, the size of the matrix grows with the number of antennas and subcarriers, resulting in an increasing amount of airtime overhead and computational load at the station. Conventional approaches come with either excessive computational load or loss of beamforming precision. For this reason, we propose SplitBeam, a new framework where we train a split deep neural network (DNN) to directly output the BM given the channel state information (CSI) matrix as input. The DNN is designed with an additional “bottleneck” layer to “split” the original DNN into a head model and a tail model, respectively executed by the station and the access point. The head model generates a compressed representation of the BM, which is then used by the AP to produce the BM using the tail model. We formulate and solve a bottleneck optimization problem (BOP) to keep computation, airtime overhead, and bit error rate (BER) below application requirements. We perform extensive experimental CSI collection with off-the-shelf Wi-Fi devices in two distinct environments and compare the performance of SplitBeam with the standard IEEE 802.11 algorithm for BM feedback and the state-of-the-art DNN-based approach LB-SciFi. Our experimental results show that SplitBeam reduces the beamforming feedback size and computational complexity by respectively up to 81 % and 84 % while maintaining BER within about 10−3of existing approaches. We also implement the SplitBeam DNNs on FPGA hardware to estimate the end-to-end BM reporting delay, and show that the latter is less than 10 milliseconds in the most complex scenario, which is the target channel sounding frequency in realistic multi-user MIMO scenarios. To allow full reproducibility, we will release our code and datasets to the community.

A key challenge of today’s systems is the mismatch between the high computational demands of modern neural network models for data analysis and the severely limited resources of mobile devices. Existing solutions focus on model simplification and task offloading to compute-capable edge servers. The former often leads to performance degradation, whereas the latter requires the transfer of information-rich signals and is subject to the impairments of wireless channels. To address these issues, a framework that establishes a novel form of collaboration between mobile devices and edge servers is proposed herein. The core idea is to deploy lightweight models on mobile devices that are intelligently updated to match the current, and local, distribution of the samples being observed. The framework develops the temporal patterns of the samples to determine the optimal model update policy, as well as channel resources allocated to the mobile users. The performance of the proposed framework is evaluated via extensive experiments with both synthetic and real-world datasets.

The execution of large deep neural networks (DNN) at mobile edge devices requires considerable consumption of critical resources, such as energy, while imposing demands on hardware capabilities. In approaches based on edge computing the execution of the models is offloaded to a compute-capable device positioned at the edge of 5G infrastructures. The main issue of the latter class of approaches is the need to transport information-rich signals over wireless links with limited and time-varying capacity. The recent split computing paradigm attempts to resolve this impasse by distributing the execution of DNN models across the layers of the systems to reduce the amount of data to be transmitted while imposing minimal computing load on mobile devices. In this context, we propose a novel split computing approach based on slimmable ensemble encoders. The key advantage of our design is the ability to adapt computational load and transmitted data size in real-time with minimal overhead and time. This is in contrast with existing approaches, where the same adaptation requires costly context switching and model loading. Moreover, our model outperforms existing solutions in terms of compression efficacy and execution time, especially in the context of weak mobile devices. We present a comprehensive comparison with the most advanced split computing solutions, as well as an experimental evaluation on GPU-less devices.

Machine learning is becoming an increasingly integral component of mobile applications. However, the execution of compute-heavy neural models (e.g., for computer vision tasks) on resource-constrained devices is challenging due to their limited computing power, memory, and energy reservoir. While edge computing mitigates these issues, the transfer of information-rich signals over capacity-limited and time-varying wireless channels may result in large latency and latency variations. Herein, we propose a methodology to route heterogeneous tasks across the resources and layers of systems composed of mobile devices and edge servers. Different from prior work, we consider aspects of real-world systems, such as context switching, task accumulation, and the interplay between communications and computing components of the overall pipeline, that are rarely captured in abstract models. To optimize the task flow, we use a deep reinforcement learning agent trained on real-world data collected using a system we developed. The agent uses an articulate definition of state drawing features from several logical blocks of the system. Results indicate that the agent adapts the routing of tasks to parameters controlling their heterogeneity, as well as the hardware setup and the state of the wireless channel.

Edge computing is to be widely adopted for autonomous systems (ASs) applications as compute-intensive processing tasks can be offloaded to compute-capable servers located at the edge of the network infrastructure. Given the critical nature of numerous AS applications, their tasks are mostly governed by strict execution deadlines to alleviate any safety concerns from delayed responses. Although wireless link uncertainty has prompted recent works to designate redundant local execution as an offloading fail-safe to ensure these deadlines are met, frequent invocation of such fail-safe mechanisms can potentially undermine the extent of performance gains from offloading. In this article, we thoroughly analyze how redundant execution overheads can influence the overall performance. Then, we present TESTUDO, a methodology to optimize the energy consumption for latency-sensitive AS applications employing collaborative edge computing. Primarily, our methodology encompasses two main stages: 1) designing processing pipelines supporting optimal offloading points and fail-safe integration using modular design techniques and 2) developing a context-aware adaptive runtime solution based on deep reinforcement learning to adapt the mode of operation according to the wireless network status. Our experiments for end-to-end control and object detection use-cases have shown that TESTUDO achieved energy gains reaching up to 31% and 13.4% (15.9% and 5.3% on average) for the former and latter, respectively, while incurring little-to-no degradation in prediction scores (< 1% change) from state-of-the-art strategies.

Cheap commercial off-the-shelf (COTS) First-Person View (FPV) drones have become widely available for consumers in recent years. Unfortunately, they also provide low-cost attack opportunities to malicious users. Thus, effective methods to detect the presence of unknown and non-cooperating drones within a restricted area are highly demanded. Approaches based on detection of drones based on emitted video stream have been proposed, but were not yet shown to work against other similar benign traffic, such as that generated by wireless security cameras. Most importantly, these approaches were not studied in the context of detecting new unprofiled drone types. In this work, we propose a novel drone detection framework, which leverages specific patterns in video traffic transmitted by drones. The patterns consist of repetitive synchronization packets (we call pivots), which we use as features for a machine learning classifier. We show that our framework can achieve up to 99% in detection accuracy over an encrypted WiFi channel using only 170 packets originated from the drone within 820ms time period. Our framework is able to identify drone transmissions even among very similar WiFi transmissions (such as video streams originated from security cameras) as well as in noisy scenarios with background traffic. Furthermore, the design of our pivot features enables the classifier to detect unprofiled drones in which the classifier has never trained on and is refined using a novel feature selection strategy that selects the features that have the discriminative power of detecting new unprofiled drones.

LiDAR is rapidly emerging as a central sensor in many applications involving autonomous navigation. However, the execution of state-of-the-art neural models for the analysis of point clouds produced by LiDARs necessitates considerable computing power, energy and memory. As a consequence, real-time analysis – e.g., 3D object detection – on resource-constrained mobile platforms such as Unmanned Aerial Vehicles (UAV) is often impractical. In this paper, we evaluate the feasibility of real-time LiDAR-based object detection for UAVs using measures and data obtained from a real-world deployment. First, we demonstrate that state-of-the-art neural models for 3D object detection cannot be executed even in relatively powerful embedded computers suitable for airborne drones, such as the NVIDIA Jetson Nano. Then, we focus our attention on edge computing, where the UAV offloads the execution of the analysis model to a compute-capable device (an edge server) positioned at the network edge. The key challenge is that point clouds generated by LiDARs have a large size (1.2MB per point cloud frame). We evaluate the overall capture-to-output delay of a remote analysis loop experimentally for WiFi and using expected data rate for cellular LTE environments. Finally, we evaluate the performance of 3D object detection on available datasets for autonomous vehicles and emphasize the challenges posed by the ability of UAVs to move in the 3D space. With our LiDAR-UAV system, we achieved detections with averages of 86% accuracy, 71.6% precision, and 55.14% recall outputted with an average end-to-end delay of 1920ms.

There has been much interest in deploying deep learning algorithms on low-powered devices, including smart-phones, drones, and medical sensors. However, full-scale deep neural networks are often too resource-intensive in terms of energy and storage. As a result, the bulk part of the machine learning operation is therefore often carried out on an edge server, where the data is compressed and transmitted. However, compressing data (such as images) leads to transmitting information irrelevant to the supervised task. Another popular approach is to split the deep network between the device and the server while compressing intermediate features. To date, however, such split computing strategies have barely outperformed the aforementioned naive data compression baselines due to their inefficient approaches to feature compression. This paper adopts ideas from knowledge distillation and neural image compression to compress intermediate feature representations more efficiently. Our supervised compression approach uses a teacher model and a student model with a stochastic bottleneck and learnable prior for entropy coding (Entropic Student). We compare our approach to various neural image and feature compression baselines in three vision tasks and found that it achieves better supervised rate-distortion performance while maintaining smaller end-to-end latency. We furthermore show that the learned feature representations can be tuned to serve multiple downstream tasks.

Wearable sensors and Internet of Things technologies are enabling automated health monitoring applications, where signals captured by sensors are analyzed in real-time by algorithms detecting health issues and conditions. However, continuous clinical-level monitoring of patients in everyday settings often requires computation, storage and connectivity capabilities beyond those possessed by wearable sensors. While edge computing partially resolves this issue by connecting the sensors to compute-capable devices positioned at the network edge, the wireless links connecting the sensors to the edge servers may not have sufficient capacity to transfer the information-rich data that characterize these applications. A possible solution is to compress the signal to be transferred, accepting the tradeoff between compression gain and detection accuracy. In this paper, we propose SIC-EDGE: a "semantic compression" framework whose goal is to dynamically optimize the resolution of an electrocardiogram (ECG) signal transferred from a wearable sensor to an edge server to perform real-time detection of heart diseases. The core idea is to establish a collaborative control loop between the sensor and the edge server to iteratively build a semantic representation that is: (i) ECG-cycle specific; (ii) personalized, and (iii) targeted to support the classification task rather than signal reconstruction. The core of SIC-EDGE is a Sequential Hypothesis Testing (SHT) algorithm that analyzes partial representations along the iterations to determine which and how many representation layers (wavelet coefficients in our implementation) are requested. Our results on established datasets demonstrates the need for adaptive "semantic" compression, and illustrate the dynamic compression strategy realized by SIC-EDGE. We show that SIC-EDGE leads to an increase in terms of recall and F1 score of up to 35% and 26% respectively compared to an optimized but static wavelet compression for a given maximum channel usage.

Mobile devices such as drones and autonomous vehicles increasingly rely on object detection (OD) through deep neural networks (DNNs) to perform critical tasks such as navigation, target-tracking and surveillance, just to name a few. Due to their high complexity, the execution of these DNNs requires excessive time and energy. Low-complexity object tracking (OT) is thus used along with OD, where the latter is periodically applied to generate "fresh" references for tracking. However, the frames processed with OD incur large delays, which does not comply with real-time applications requirements. Offloading OD to edge servers can mitigate this issue, but existing work focuses on the optimization of the offloading process in systems where the wireless channel has a very large capacity. Herein, we consider systems with constrained and erratic channel capacity, and establish parallel OT (at the mobile device) and OD (at the edge server) processes that are resilient to large OD latency. We propose Katch-Up, a novel tracking mechanism that improves the system resilience to excessive OD delay. We show that this technique greatly improves the quality of the reference available to tracking, and boosts performance up to 33%. However, while Katch-Up significantly improves performance, it also increases the computing load of the mobile device. Hence, we design SmartDet, a low-complexity controller based on deep reinforcement learning (DRL) that learns to achieve the right trade-off between resource utilization and OD performance. SmartDet takes as input highly-heterogeneous context-related information related to the current video content and the current network conditions to optimize frequency and type of OD offloading, as well as Katch-Up utilization. We extensively evaluate SmartDet on a real-world testbed composed by a JetSon Nano as mobile device and a GTX 980 Ti as edge server, connected through a Wi-Fi link, to collect several network-related traces, as well as energy measurements. We consider a state-of-the-art video dataset (ILSVRC 2015 - VID) and state-of-the-art OD models (EfficientDet 0, 2 and 4). Experimental results show that SmartDet achieves an optimal balance between tracking performance – mean Average Recall (mAR) and resource usage. With respect to a baseline with full Katch-Up usage and maximum channel usage, we still increase mAR by 4% while using 50% less of the channel and 30% power resources associated with Katch-Up. With respect to a fixed strategy using minimal resources, we increase mAR by 20% while using Katch-Up on 1/3 of the frames.

Although mission-critical applications require the use of deep neural networks (DNNs), their continuous execution at mobile devices results in a significant increase in energy consumption. While edge offloading can decrease energy consumption, erratic patterns in channel quality, network and edge server load can lead to severe disruption of the system’s key operations. An alternative approach, called split computing, generates compressed representations within the model (called "bottlenecks"), to reduce bandwidth usage and energy consumption. Prior work has proposed approaches that introduce additional layers, to the detriment of energy consumption and latency. For this reason, we propose a new framework called BottleFit, which, in addition to targeted DNN architecture modifications, includes a novel training strategy to achieve high accuracy even with strong compression rates. We apply BottleFit on cutting-edge DNN models in image classification, and show that BottleFit achieves 77.1% data compression with up to 0.6% accuracy loss on ImageNet dataset, while state of the art such as SPINN loses up to 6% in accuracy. We experimentally measure the power consumption and latency of an image classification application running on an NVIDIA Jetson Nano board (GPU-based) and a Raspberry PI board (GPU-less). We show that BottleFit decreases power consumption and latency respectively by up to 49% and 89% with respect to (w.r.t.) local computing and by 37% and 55% w.r.t. edge offloading. We also compare BottleFit with state-of-the-art autoencoders-based approaches, and show that (i) BottleFit reduces power consumption and execution time respectively by up to 54% and 44% on the Jetson and 40% and 62% on Raspberry PI; (ii) the size of the head model executed on the mobile device is 83 times smaller. We publish the code repository for reproducibility of the results in this study.

Edge computing enables Mobile Autonomous Systems (MASs) to execute continuous streams of heavy-duty mission-critical processing tasks, such as real-time obstacle detection and navigation. However, in practical applications, erratic patterns in channel quality, network load, and edge server load can interrupt the task flow's execution, which necessarily leads to severe disruption of the system's key operations. Existing work has mostly tackled the problem with reactive approaches, which cannot guarantee task-level reliability. Conversely, in this paper we focus on learning-based predictive edge computing to achieve self-resilient task offloading. By conducting a preliminary experimental evaluation, we show that there is no dominant feature that can predict the edge-MAS system reliability, which calls for an ensemble and selection of weaker features. To tackle the complexity of the problem, we propose SeReMAS, a data-driven optimization framework. We first mathematically formulate a Redundant Task Offloading Problem (RTOP), where a MAS may connect to multiple edge servers for redundancy, and needs to select which server(s) to transmit its computing tasks in order to maximize the probability of task execution while minimizing channel and edge resource utilization. We then create a predictor based on Deep Reinforcement Learning (DRL), which produces the optimum task assignment based on application-, network- and telemetry-based features. We prototype SeReMAS on a testbed composed by a Tarot650 quadcopter drone, mounting a PixHawk flight controller, a Jetson Nano board, and three 802.11n WiFi interfaces. We extensively evaluate SeReMAS by considering an application where one drone offloads high-resolution images for real-time analysis to three edge servers on the ground. Experimental results show that SeReMAS improves the task execution probability by 17% with respect to existing reactive-based approaches. To allow full reproducibility of results, we share the dataset and code with the research community.

Infrastructure assistance has been proposed as a viable solution to improve the capabilities of commercial Unmanned Aerial Vehicles (UAV), especially toward fully autonomous operations. The airborne nature of these devices imposes constrains limiting the onboard available energy supply and computing power. The assistance of the surrounding communication and computing infrastructure can mitigate such limitations by extending the communication range and taking over the execution of compute-intense tasks. However, autonomous operations impose specific, and rather extreme in some cases, demands to the infrastructure. Focusing on flight assistance and task offloading to edge servers, this paper presents an in-depth evaluation of the ability of the communication infrastructure to support the necessary flow of information from the UAV to the infrastructure. The study is based on our recently proposed FlyNetSim, an open-source UAV-network simulator accurately modeling both UAV and network operations.

Many modern applications rely on complex machine learning algorithms, such as Deep Neural Networks (DNNs), to analyze images. However, both mobile and edge computing strategies may fail to provide satisfactory performance in some parameter regions. To mitigate this issue, the research community recently proposed methods to split the execution of DNNs to optimize the balance between computing load allocation and channel usage. Building on this set of results, this paper presents an optimization framework that enables the dynamic control of how images are processed in mobile device-edge server systems. The system is modeled as a Markov process, and a Linear Fractional Program is defined to identify the optimal stationary state-action distribution minimizing the overall average inference time under a constraint on the number of discarded images. Results indicate the advantage of using a dynamic control strategy with respect to available fixed strategies.

With the ever-increasing popularity of fitness trackers, data on the time and location of popular walking, running, and bicycling routes is expansive and growing rapidly. This data is currently used primarily for route discovery and personal fitness tracking, but it may also be leveraged to build ad-hoc transportation flows. We present a novel model that creates delivery networks from these zero-emission transportation flows, and we evaluate the model using data from two popular datasets. Our results indicate that such networks are indeed possible, and can help reduce traffic, emissions, and delivery times. Moreover, we demonstrate how our results can be consistently reproduced in different cities with different subsets of carriers.

The analysis of information rich signals is at the core of autonomy. In airborne devices such as Unmanned Aerial Vehicles (UAV), the hardware limitations imposed by the weight constraints make the continuous execution of these algorithms challenging. Edge computing can mitigate such limitations and boost the system and mission performance of the UAVs. However, due to the UAVs motion characteristics and complex dynamics of urban environments, remote processing-control loops can quickly degrade. This paper presents Hydra, a framework for the dynamic selection of communication/computation resources in this challenging environment. A full - open-source - implementation of Hydra is discussed and tested via real-world experiments.

The edge computing paradigm places compute-capable devices - edge servers - at the network edge to assist mobile devices in executing data analysis tasks. Intuitively, offloading compute-intense tasks to edge servers can reduce their execution time. However, poor conditions of the wireless channel connecting the mobile devices to the edge servers may degrade the overall capture-to-output delay achieved by edge offloading. Herein, we focus on edge computing supporting remote object detection by means of Deep Neural Networks (DNNs), and develop a framework to reduce the amount of data transmitted over the wireless link. The core idea we propose builds on recent approaches splitting DNNs into sections - namely head and tail models - executed by the mobile device and edge server, respectively. The wireless link, then, is used to transport the output of the last layer of the head model to the edge server, instead of the DNN input. Most prior work focuses on classification tasks and leaves the DNN structure unaltered. Herein, our focus is on DNNs for three different object detection tasks, which present a much more convoluted structure, and modify the architecture of the network to: (i) achieve in-network compression by introducing a bottleneck layer in the early layers on the head model, and (ii) prefilter pictures that do not contain objects of interest using a convolutional neural network. Results show that the proposed technique represents an effective intermediate option between local and edge computing in a parameter region where these extreme point solutions fail to provide satisfactory performance. The code and trained models are available at https://github.com/yoshitomo-matsubaralhnd-ghnd-object-detectors.

Healthcare applications supported by the Internet of Things enable personalized monitoring of a patient in everyday settings. Such applications often consist of battery-powered sensors coupled to smart gateways at the edge layer. Smart gateways offer several local computing and storage services (e.g., data aggregation, compression, local decision making), and also provide an opportunity for implementing local closed-loop optimization of different parameters of the sensor layer, particularly energy consumption. To implement efficient optimization methods, information regarding the context and state of patients need to be considered to find opportunities to adjust energy to demanded accuracy. Edge-assisted optimization can manage energy consumption of the sensor layer but may also adversely affect the quality of sensed data, which could compromise the reliable detection of health deterioration risk factors. In this article, we propose two approaches: myopic and Markov decision processes (MDPs)—to consider both energy constraints and risk factor requirements for achieving a twofold goal: energy savings while satisfying accuracy requirements of abnormality detection in a patient’s vital signs. Vital signs, including heart rate, respiration rate, and oxygen saturation, are extracted from a photoplethysmogram signal and errors of extracted features are compared to a ground truth that is modeled as a Gaussian distribution. We control the sensor’s sensing energy to minimize the power consumption while meeting a desired level of satisfactory detection performance. We present experimental results on realistic case studies using a reconfigurable photoplethysmogram sensor in an IoT system, and show that compared to nonadaptive methods, myopic reduces an average of 16.9% in sensing energy consumption with the maximum probability of abnormality misdetection on the order of 0.17 in a 24-hour health monitoring system. In addition, over 4 weeks of monitoring, we demonstrate that our MDP policy can extend the battery life on average of more than 2x while fulfilling the same average probability of misdetection compared to the myopic method. We illustrate results comparing myopic , MDP, and nonadaptive methods to monitor 14 subjects over 1 month.

Recent advances in pervasive Internet of Things technologies and edge computing have opened new avenues for development of ubiquitous health monitoring applications. Delivering an acceptable level of usability and accuracy for these healthcare Internet of Things applications requires optimization of both system-driven and data-driven aspects, which are typically done in a disjoint manner. Although decoupled optimization of these processes yields local optima at each level, synergistic coupling of the system and data levels can lead to a holistic solution opening new opportunities for optimization. In this article, we present an edge-assisted resource manager that dynamically controls the fidelity and duration of sensing w.r.t. changes in the patient’s activity and health state, thus fine-tuning the trade-off between energy efficiency and measurement accuracy. The cornerstone of our proposed solution is an intelligent low-latency real-time controller implemented at the edge layer that detects abnormalities in the patient’s condition and accordingly adjusts the sensing parameters of a reconfigurable wireless sensor node. We assess the efficiency of our proposed system via a case study of the photoplethysmography-based medical early warning score system. Our experiments on a real full hardware-software early warning score system reveal up to 49% power savings while maintaining the accuracy of the sensory data.

In the latest years, the popularity of commercial drones has grown rapidly due to their cheaper costs and great availability on the market. The great diffusion of remotely piloted devices unfortunately leads to several security and safety concerns that need to be tackled. In this paper, we consider a fingerprint-based drone detection approach relying on the analysis of WiFi traffic features to identify the presence of unauthorized devices. In particular, we study the statistical distribution of the features composing the fingerprint vector, and we propose an adversarial approach to drone detection in order to invalidate the reliability of the surveillance system, by introducing fake ad-hoc traffic features. Results show that our novel approach is able to deceive the drone detection system through the introduction of flows belonging to arbitrary traffic classes. Also, the proposed adversarial method provides the expected significant impact on the performance of the system, actually reducing the recognition accuracy to about 50%.

The quasi-periodic nature of electrocardiogram (ECG) signals enables the use of compression techniques to minimize communications and reduce energy intake for diagnostic and preventive health monitoring. However, compression often degrades signal quality and may impair analysis by means of machine learning algorithms for the detection of anomalies. In this paper, we present an approach to pre-select relevant portions of the ECG signal at the sensor to reduce network load while satisfying a predefined diagnostic sensitivity requirement. We deploy a Deep Neural Network (DNN) to filter-out the signal's normal rhythms and reduce the amount of data stored or transmitted for further processing. Our extensive experiments covering a wide range of DNN hyper-parameters illustrate the tradeoff between diagnostic sensitivity, channel usage, energy consumption and computational complexity.

In this paper, we propose an equivalent channel model for the analysis of passive backscattering systems (e.g., Radio Frequency Identification systems). The proposed framework accurately models radiation efficiencies at the backscattering nodes whose communication channels are affected by spatially correlated shadowing. First, we derive the distribution of interference in the system and the probability that a node will activate as a function of a specific geographical distribution of the nodes. Then, we approximate the capture probability using a multivariate moment matching approach. The rationale is to provide an underlying structure for the cross-layer analysis of current MAC protocols with the perspective of performance enhancement. Numerical results illustrate the performance of a standard MAC protocol for passive RFID systems, including an accurate evaluation of the impact of the channel characterization.

The combination of computation and communication constraints within the Internet of Things systems require intelligent allocation of decision making and learning processes across a network of sensing and computing devices. In this paper, we present the problem of observation selection for reactive on-sensor decision-making, where the most accurate decision rule cannot be used unaided neither at the sensor (due to limited computing power), nor in the cloud (due to high communication latency). To make time-sensitive adaptation possible in these conditions, we consider learning a decision rule that is computationally viable for on-sensor use and is continuously adjusted by the cloud using the optimal decision rule for supervision. We pose a constrained stochastic optimization problem for online learning of such instrumental on-sensor classifier, propose an algorithm for updating its parameters, and establish the conditions under which convergence to a local extremum is guaranteed, at least for samples of independent observations.

Texting while Driving has been reported as one of the major sources of inattention by car drivers, leading to an increased probability of severe road accidents. In fact, notifications, messages and other interactions with mobile devices may make the driver unaware of road and traffic events. To prevent or mitigate this issue, solutions have been proposed that either block the smartphone when inside the vehicle or recognize the activity to issue monetary fines at a later time. This paper proposes a classification framework capable to identify the location of a device within the vehicle using data from integrated sensors. This allow more selective countermeasures targeted specifically to mobile devices used by the driver, rather than by any person inside the vehicle. The framework extracts sensor data from the smartphone, computes ad-hoc features and feeds them to a neural network. Different from prior work, we demonstrate that accurate detection can be achieved even using only one device by combining subsequent turns of the vehicle.

The Internet of Things is a key enabler of mobile health-care applications. However, the inherent constraints of mobile devices, such as limited availability of energy, can impair their ability to produce accurate data and, in turn, degrade the output of algorithms processing them in real-time to evaluate the patient's state. This paper presents an edge-assisted framework, where models and control generated by an edge server inform the sensing parameters of mobile sensors. The objective is to maximize the probability that anomalies in the collected signals are detected over extensive periods of time under battery-imposed constraints. Although the proposed concept is general, the control framework is made specific to a use-case where vital signs -heart rate, respiration rate and oxygen saturation- are extracted from a Photoplethysmogram (PPG) signal to detect anomalies in real-time. Experimental results show a 16.9% reduction in sensing energy consumption in comparison to a constant energy consumption with the maximum misdetection probability of 0.17 in a 24-hour health monitoring system.

Context-aware computing uses the wide range of information produced by mobile platforms to optimize the parameters of applications providing important services. However, as some of the data are either publicly exposed or can be acquired by malicious parties, a privacy issue arises. Recent studies extend this concept to context prediction, that is, the ability to forecast the future user context from current or past data. Whereas privacy in context-aware computing has been widely studied, the issue of impairing the ability to predict user context remains largely unexplored. This paper presents a framework based on Markov process theory to reduce the accuracy of predictions made by a malicious observer. Rather than attempting to hide current context, which is often purposely exposed by the user, the proposed methodology injects manufactured, and temporary, context updates to impair prediction. Numerical results from FourSquare databases demonstrate the privacy increase granted by the proposed technique and illustrate the tradeoff between user privacy and noise injection.

The ability of Unmanned Aerial Vehicles (UAV) to autonomously operate is constrained by the severe limitations of on-board resources. The limited processing speed and energy storage of these devices inevitably makes the real-time analysis of complex signals the key to autonomy challenging. In urban environments, the UAV can leverage the communication and computation resources of the surrounding city-wide Internet of Things infrastructure to enhance their capabilities. For instance, the UAVs can interconnect with edge computing resources and offload computation task to improve response time to sensor input and reduce energy consumption. However, the complexity of the urban topology and the large number of devices and data streams competing for the same network and computation resources create an extremely dynamic environment, where poor channel conditions and edge server congestion may penalize the performance of task offloading. This paper develops a framework enabling optimal offloading decisions as a function of network and computation load parameters and current state. The optimization is formulated as an optimal stopping time problem over a Markov process.

Unmanned Aerial Vehicle (UAV) systems are being increasingly used in a broad range of applications requiring extensive communications, either to interconnect the UAVs with each other or with ground resources. Focusing either on the modeling of UAV operations or communication and network dynamics, available simulation tools fail to capture the complex interdependencies between these two aspects of the problem. The main contribution of this paper is a flexible and scalable open source simulator -- FlyNetSim -- bridging the two domains. The overall objective is to enable simulation and evaluation of UAV swarms operating within articulated multi-layered technological ecosystems, such as the Urban Internet of Things (IoT). To this aim, FlyNetSim interfaces two open source tools, ArduPilot and ns-3, creating individual data paths between the devices operating in the system using a publish and subscribe-based middleware. The capabilities of FlyNetSim are illustrated through several case-study scenarios including UAVs interconnecting with a multi-technology communication infrastructure and intra-swarm ad-hoc communications.

Indoor positioning systems consist of identifying the physical location of devices inside buildings. They are usually based on the signal strength of a device packet received by a set of WiFi access points. Among the most precise solutions, are those based on machine learning algorithms, such as kNN (k-Nearest Neighbors). This technique is known as fingerprint positioning. Even though kNN is one of the most used classification methods due to its high precision results, it lacks scalability since an instance we need to classify must be compared to all other instances in the training base. In this work, we use a novel hierarchical navigable small world graph technique to fit the training database so that the samples can be efficiently classified in the online phase of the fingerprint positioning, allowing it to be used in large-scale scenarios and/or to be executed in resource-limited devices. We evaluated the performance of this solution using both synthetic and real-world training data and compared its performance to other known kNN variants such as kd-tree and ball-tree. Our results clearly show the performance gains of the graph-based solution, while still being able to maintain or even reduce the positioning error.

A two-hop relay network is considered, in which an eavesdropper can overhear the relaying signal. To prevent the eavesdropper from decoding this signal, a destination transmits a jamming noise while a source transmits the data signal to the relay. At the same time, the relay can harvest energy from both the source signal and the jamming noise, and use this harvested energy to forward the received signal to the destination. In such a wireless-powered relay system with cooperative jamming, we propose two adaptive relaying schemes based on power splitting and time switching techniques. In the proposed power splitting-based relaying (PSR) and time switching-based relaying (TSR) schemes, the relay controls the power splitting ratio (p) and time switching ratio (α), respectively, in order to achieve a balance between signal processing and energy harvesting. We find analytically the optimal values of p and α in each scheme to maximize the secrecy capacity under the assumption of high signal-to-noise ratio (SNR). Interestingly, although the eavesdropper's channel state information (CSI) is used in the derivation of the optimal control parameters (p and α), they are shown not to be affected by the eavesdropper's CSI in a high SNR regime. This implies that the proposed schemes can be effective even for practical environments where there is no eavesdropper's CSI. Furthermore, simulation results show that they well coincide with the exact solutions in practical environments even though the closed-form solutions are obtained with a high SNR assumption. Moreover, the comparisons of PSR and TSR in various scenarios show that the two relaying schemes have complementary performances depending on the network conditions. Specifically, PSR achieves greater secrecy capacity than TSR when the channel condition is unfavorable to the eavesdropper for wiretapping.

Energy harvesting is a key technology for future energy neutral ubiquitous wireless communication systems. Among the many available sources of energy, RF harvesting is a natural choice for wireless communications since the devices can be efficiently designed for RF harvesting operations. To achieve energy autonomy a rechargeable energy storage is needed. However, the bursty nature of energy arrival associated with wireless power transfer and the energy usage pattern induced by devices' activity may cause considerable stress to the battery and reduce its life span. In fact, deep charging and discharging cycles degrade the battery state of health, that is, the maximum amount of energy that can be stored. In this paper, a framework for the optimization of wireless nodes' transmission and charging strategy is presented, where the battery aging rate is made explicit as a performance metric. The proposed framework is based on MDP theory, where the embedded stochastic model captures the energy arrival and data generation processes, and the control variable takes the form of a transmission power selection and energy acceptance/rejection. The tradeoff between the achievable throughput and the aging rate is explored for different parameters setting.

This paper develops a control framework for a network of energy harvesting nodes connected to a Base Station (BS) over a multiple access channel. Due to fluctuations in energy availability and, possibly, energy outages, the number of nodes attempting channel access is random and varies over time. Thus, each node must carefully adapt its access probability to the network state to optimize network performance. In order to reduce the complexity of network control, a lightweight and flexible design framework is proposed where energy storage dynamics are replaced by dynamic average power constraints at each node, induced by the time correlated energy supply. The BS adapts the access probability of the “active” nodes (those currently under a favorable energy harvesting state) so as to maximize throughput. The resulting policy takes the form of access probability as a function of the local energy harvesting state and number of active nodes. The structure of the throughput-optimal policy is analytically derived for the genie-aided case of non-causal knowledge of the number of active nodes. Inspired by it, a Bayesian estimation approach is presented for the more practical scenario where the BS estimates the number of active nodes. The proposed scheme is shown to outperform by 20% a scheme in which the nodes operate based only on local state information, and to be robust against the impact of energy storage dynamics at a fraction of the complexity.

By placing processing-capable devices at the edge of local wireless access networks, Edge Computing architectures have been recently proposed to connect mobile devices to computational power through a one-hop low-latency wireless link. In this paper, we propose a new design where edge assistance is used to control local data filtering at the mobile devices in bandwidth and energy constrained systems. We focus on real-time monitoring applications, where the video input from mobile devices is processed to centrally detect and recognize objects. The edge processor controls the activation and deactivation of local classifiers implemented by the mobile devices to remove useless portions of video frames. The objective is to adapt the video stream to time-varying bandwidth constraints, while minimizing the additional energy consumption introduced by local processing. To this end, an optimization problem is formulated for a loss function embodying the balance between the risk of violating the available bandwidth and the cost of overly-conservative data filtering. The edge assists the local decision by extracting parameters of the video, such as density of objects of interest in a frame, which influence the output of the sensor. Numerical results, obtained by performing a measurement campaign based on a real implementation, illustrate the tension between energy and bandwidth use for a Haar-feature based cascade classifier.

A novel approach to cognitive networks is proposed, where cognition is used to optimize the delivery of content in wireless networks. Whereas most prior work in this area divides users into primary and secondary users, the proposed approach defines content classes with different quality of service requirements. The cognitive terminals implement smart channel access control strategies to enable the coexistence of heterogeneous applications on the same network resource. A Markov Decision Process framework is built to derive and investigate the structure of the optimal control policy in the form of action distribution over the network state space. For the considered problem, the optimal policy presents a specific structure that is analytically proven by means of renewal process analysis.

The Internet of Things (IoT) paradigm holds significant promises for remote health monitoring systems. Due to their life- or mission-critical nature, these systems need to provide a high level of availability and accuracy. On the one hand, centralized cloud-based IoT systems lack reliability, punctuality and availability (e.g., in case of slow or unreliable Internet connection), and on the other hand, fully outsourcing data analytics to the edge of the network can result in diminished level of accuracy and adaptability due to the limited computational capacity in edge nodes. In this paper, we tackle these issues by proposing a hierarchical computing architecture, HiCH, for IoT-based health monitoring systems. The core components of the proposed system are 1) a novel computing architecture suitable for hierarchical partitioning and execution of machine learning based data analytics, 2) a closed-loop management technique capable of autonomous system adjustments with respect to patient’s condition. HiCH benefits from the features offered by both fog and cloud computing and introduces a tailored management methodology for healthcare IoT systems. We demonstrate the efficacy of HiCH via a comprehensive performance assessment and evaluation on a continuous remote health monitoring case study focusing on arrhythmia detection for patients suffering from CardioVascular Diseases (CVDs).

In the Urban Internet of Things (IoT), devices and systems are interconnected at the city scale to provide innovative services to the citizens. However, the traffic generated by the sensing and processing systems may overload local access networks. A coexistence problem arises where concurrent applications mutually interfere and compete for available resources. This effect is further aggravated by the multiple scales involved and heterogeneity of the networks supporting the urban IoT. One of the main contributions of this paper is the introduction of the notion of content- oriented cognitive interference control in heterogeneous local access networks supporting computing and data processing in the urban IoT. A network scenario where multiple communication technologies, such as Device-to- Device and Long Term Evolution (LTE), is considered. The focus of the present paper is on city monitoring applications, where a video data stream generated by a camera system is remotely processed to detect objects. The cognitive network paradigm is extended to dynamically shape the interference pattern generated by concurrent data streams and induce a packet loss trajectory compatible with video processing algorithms. Numerical results show that the proposed cognitive transmission strategy enables a significant throughput increase of interfering applications for a target accuracy of the monitoring application.

Energy harvesting is arising as a key technology in wireless systems, allowing continuous and prolonged operations. However, the bursty nature of the energy arrival process associated with renewable sources and the energy usage pattern caused by wireless protocols may cause considerable stress to the battery and eventually reduce its lifetime. In fact, deep charging and discharging cycles degrade the battery State of Health, that is, the maximum amount of energy that can be stored. In this paper, a framework for the optimization of wireless nodes' transmission strategy is presented, where battery aging rate is included as a constraint. The proposed framework is based on Markov Decision Process theory, where the embedded stochastic process models energy arrival and storage, and channel fading, as well as the control variables. Numerical results unveil the tension between packet delivery rate and battery degradation.

Energy harvesting from environments such as solar energy are promising solutions to tackle energy sustainability in embedded systems. However, uncertainties in energy availability, non-ideal characteristics of harvesting circuits, energy storage (battery or supercapacitor), and application demand dynamics add more complexity in the system. We present a unified model based on discrete-time Finite State Markov Chain to capture the dynamicity and variations in both the energy supply from solar irradiance and the energy demand from the application. In this paper, we exploit the temporal and spatial characteristics of solar energy and propose a deterministic profile with stochastic process to reflect the fluctuation due to unexpected weather condition. Optimal policy to maximize expected total QoS is derived from the presented model using a probabilistic dynamic programming approach. Compared to a state-of-the-art deterministic energy management framework, our proposed approach outperforms in term of QoS and energy sustainability (with less shutdown time) of the system.

A framework for the distributed optimization of reactive cognitive networks with multiple secondary users is presented. The secondary users iteratively locate the policy maximizing their aggregate performance under bounds on the maximum performance loss caused to the primary users. The policy accounts for the impact of interference on the dynamics of the primary users' network due to transmission and channel access protocols. To minimize coordination overhead, it is assumed that the secondary users only coordinate the policy, whereas actions in each slot are independently selected by the individual secondary user based on the agreed policy. The distributed optimization technique proposed herein is based on alternating convex optimization. Numerical results are presented assessing the performance of the obtained transmission policy with respect to the optimal centralized and fully-coordinated policy.

ABSTRACT A novel iterative algorithm for the efficient computation of the intersection areas of an arbitrary number of circles is presented. The algorithm, on the basis of a trellis structure, hinges on two geometric results, which allow the existence check and the computation of the area of the intersection regions generated by more than three circles by simple algebraic manipulations of the intersection areas of a smaller number of circles. The presented algorithm is a powerful tool for the performance analysis of wireless networks and finds many applications, ranging from sensor to cellular networks. As an example of practical application, an insightful study of the uplink outage probability in a wireless network with cooperative access points as a function of the transmission power and access point density is presented. Copyright © 2012 John Wiley & Sons, Ltd.

In this paper, the problem of state tracking with controlled observations is considered for a system modeled by a discrete-time, finite-state Markov chain. The system state is `hidden' and observed via conditionally Gaussian measurements that are shaped by the underlying state and an exogenous control input. Following an innovations approach, a Kalman-like filter is derived to estimate the Markov chain system state. To optimize the control strategy, the associated mean-squared error is used as an optimization criterion for a partially observable Markov Decision Process (POMDP). The optimal solution is determined via stochastic dynamic programming. Numerical results are presented for the application of physical activity detection in heterogeneous, wireless body area networks.

The problem of enhancing the quality of system state estimates is considered for a special class of dynamical systems. Specifically, a system characterized by a discrete-time, finite-state Markov chain state and observed via conditionally Gaussian measurements is assumed. The associated mean vectors and covariance matrices are tightly intertwined with the system state and a control input selected by a controller. Exploiting an innovations approach, finite-dimensional, non-linear approximate MMSE smoothing estimators are derived for the Markov chain system state. The resulting smoothers are driven by a control policy determined by a stochastic dynamic programming algorithm, which minimizes the MSE filtering error, and was proposed in our earlier work. An application of the smoothers derived in this paper is presented for the problem of physical activity detection in wireless body sensing networks, which illustrates the performance enhancement due to smoothing.

This paper introduces a novel technique for access by a cognitive Secondary User (SU) using best-effort transmission to a spectrum with an incumbent Primary User (PU), which uses Type-I Hybrid ARQ. The technique leverages the primary ARQ protocol to perform Interference Cancellation (IC) at the SU receiver (SUrx). Two IC mechanisms that work in concert are introduced: Forward IC, where SUrx, after decoding the PU message, cancels its interference in the (possible) following PU retransmissions of the same message, to improve the SU throughput; Backward IC, where SUrx performs IC on previous SU transmissions, whose decoding failed due to severe PU interference. Secondary access policies are designed that determine the secondary access probability in each state of the network so as to maximize the average long-term SU throughput by opportunistically leveraging IC, while causing bounded average long-term PU throughput degradation and SU power expenditure. It is proved that the optimal policy prescribes that the SU prioritizes its access in the states where SUrx knows the PU message, thus enabling IC. An algorithm is provided to optimally allocate additional secondary access opportunities in the states where the PU message is unknown. Numerical results are shown to assess the throughput gain provided by the proposed techniques.

A novel framework for the analysis and optimization of wireless networks operations is proposed. The temporal evolution of the state of the network is modeled as the trajectory of the state of a Finite State Machine (FSM). The state space of the FSM and the statistics of state transition are represented as a directed graph. Graph reduction and transform techniques are proposed to reduce the dimension of the graph associated with the FSM and analyze the properties of functions defined on its state space. The proposed methodology is based on the intrinsic multi-dimensional/multi-scale structure of the state space of the FSM and enables the analysis and minimization of cost-to-go functions, i.e., functions measuring the expected long-term cost associated with a control strategy, on coarser versions of the original graph.

Wireless Body Area Networks (WBANs) refer to a class of wireless sensor networks that are expected to support a wide variety of applications ranging from healthcare and emergency response to entertainment and sports. A WBAN can be characterized by a small number of heterogeneous sensors and an energy-constrained fusion center e.g. cellphone. A key goal is to maximize the lifetime of such a unique sensor network and based on an actual implementation of a prototype WBAN, the limited energy budget of the fusion center is a critical impediment. To overcome this issue, the stochastic control framework introduced in our earlier work is extended to account for less number of samples and sensor heterogeneity is redefined in terms of worst-case detection error probability. A Maximum Likelihood detector based on the belief state is also introduced to increase the detection performance. To account for the energy-constrained fusion center, our initial optimization problem is reformulated to two alternative, but distinct constrained versions and two completely new algorithms, E2MBADP and GME2PS2, are devised. Simulations on real-world data are provided to validate the schemes' performance. Energy gains on the order of 64% while achieving the same detection accuracy (99%) as an equal allocation scheme across sensors are observed.

Wireless body area networks (WBANs) are emerging as a powerful tool for health management, emergency response, military personnel wellness as well as sports and entertainment. In contrast to traditional sensor networks for, say, environmental sensing, WBANs are often characterized by a modest number of heterogeneous sensors wirelessly coupled to a fusion center such as a mobile phone. Based on an actual implementation of a prototype WBAN, energy efficiency at the fusion center has proven to be one of the critical roadblocks to long-term deployment of WBANs. To this end, a novel formulation based on stochastic control tools is devised to model the sensor selection process. Sensors are heterogeneous both in their discrimination capabilities as well as their energy cost, further challenging sensor selection. The goal is to maximize the WBAN's lifetime while optimizing the performance of a physical state detection application. To this end, an optimal dynamic programming algorithm is derived. However, due to the prohibitive complexity of the optimal method, a low-cost approximation scheme, T3S, is designed. The low complexity design is based on several key properties of the cost functional. The proposed T3S scheme is evaluated on real-world data collected from an implemented WBAN and observed to offer near optimal performance with significantly lower complexity.

Cooperative techniques have been shown to significantly improve the performance of wireless networks. Despite being a mature technology in single communication link scenarios, their implementation in wider, and practical, networks poses several challenges which have not been fully identified and understood so far. In this paper, the implementation of cooperative communications in non-centralized ad hoc networks with sensing-based channel access is extensively discussed. Both analysis and simulation are employed to provide a clear understanding of the mutual influence between the link layer contention mechanism and relaying protocols. Results show that sensing-based channel access significantly hampers the effectiveness of cooperation by biasing the spatial distribution of available relays, and by inducing a level of spatial and temporal correlation of interference that diminishes the diversity improvement on which cooperative gains are founded. Moreover, the efficiency reduction entailed by several practical protocol issues related to carrier sense multiple access which are typically neglected in the literature is thoroughly investigated.

Cognitive radio methodologies have the potential to dramatically increase the throughput of wireless systems. Herein, control strategies which enable the superposition in time and frequency of primary and secondary user transmissions are explored in contrast to more traditional sensing approaches which only allow the secondary user to transmit when the primary user is idle. In this paper, the optimal transmission policy for the secondary user when the primary user adopts a retransmission-based error control scheme is investigated. The policy aims to maximize the secondary users' throughput, with a constraint on the throughput loss and failure probability of the primary user. Due to the constraint, the optimal policy is randomized, and determines how often the secondary user transmits according to the retransmission state of the packet being served by the primary user. The resulting optimal strategy of the secondary user is proven to have a unique structure. In particular, the optimal throughput is achieved by the secondary user by concentrating its transmission, and thus its interference to the primary user, in the first transmissions of a primary user packet. The rather simple framework considered in this paper highlights two fundamental aspects of cognitive networks that have not been covered so far: 1) the networking mechanisms implemented by the primary users (error control by means of retransmissions in the considered model) react to secondary users' activity; 2) if networking mechanisms are considered, then their state must be taken into account when optimizing secondary users' strategy, i.e., a strategy based on a binary active/idle perception of the primary users' state is suboptimal.

An algorithm for the optimization of secondary user's transmission strategies in cognitive networks with imperfect network state observations is proposed. The secondary user minimizes the time average of a cost function while generating a bounded performance loss to the primary users' network. The state of the primary users' network, defined as a collection of variables describing features of the network (e.g., buffer state, ARQ state) evolves over time according to a homogeneous Markov process. The statistics of the Markov process is dependent on the strategy of the secondary user and, thus, the instantaneous idleness/transmission action of the secondary user has a long-term impact on the temporal evolution of the network. The Markov process generates a sequence of states in the state space of the network that projects onto a sequence of observations in the observation space, that is, the collection of all the observations of the secondary user. Based on the sequence of observations, the proposed algorithm iteratively optimizes the strategy of the secondary users with no a priori knowledge of the statistics of the Markov process and of the state-observation probability map.

An algorithm for the optimization of secondary user's transmission strategies in cognitive networks with imperfect network state observations is presented. The task of the secondary user is to maximize its performance while generating a bounded performance loss to the primary users' network. The state of the primary users' network, defined as a collection of variables describing features of the network (e.g., buffer state, ARQ state), evolves according to a Markov process whose statistics depend on the transmission strategy of the secondary user. The main contribution of this paper is an online learning algorithm that, without any a priori knowledge about the statistics of the network and state-observation map, iteratively optimizes the strategy of the secondary user based on a sample-path of noisy and partial state observations.

We study optimal transmission strategies in interfering wireless networks, under Quality of Service constraints. A buffered, dynamic network with multiple sources is considered, and sources use a retransmission strategy in order to improve packet delivery probability. The optimization problem is formulated as a Markov Decision Process, where constraints and objective functions are ratios of time-averaged cost functions. The optimal strategy is found as the solution of a Linear Fractional Program, where the optimization variables are the steady-state probability of state-action pairs. Numerical results illustrate the dependence of optimal transmission/interference strategies on the constraints imposed on the network.

In this paper, we address the problem of packet forwarding in multi-hop wireless networks of nodes equipped with multiple antennas. Through the use of MIMO (multiple-input multiple-output) technology, the physical layer is able to superimpose multiple concurrent information flows, thereby increasing the communications parallelism in the network. We propose a cooperative cross-layer scheme that integrates distributed incremental redundancy hybrid automatic retransmission request (HARQ) error control with routing in order to increase the efficiency of the transmissions and decrease the interference load in the network. In the proposed scheme, next hop relays are opportunistically selected during the HARQ process in order to avoid overloaded receivers and guarantee the geographical advancement of the packets. Thus, cooperation is used to both strengthen weak links and select the path to the destination. Numerical results are presented for a specific instantiation where a layered space-time multiuser detection (LASTMUD) transmitter-receiver architecture with spread-spectrum signals is implemented at the physical layer and packets are encoded with a linear erasure code (LEC).

This paper presents performance analysis and cross-layer design approaches for hybrid ARQ (HARQ) protocols in wireless networks, which employ adaptive modulation and coding (AMC) in conjunction with adaptive cooperative diversity and are subject to time-correlated fading channels. We first consider a point-to-point scenario, i.e., non-cooperative HARQ with AMC. Utilizing a Markov channel model which accounts for the temporal correlation in the successive transmission of incremental redundancy by the HARQ protocol, we derive the system throughput and the packet loss probability based on a rate compatible punctured convolutional code family. Next, we consider a cooperative HARQ (CHARQ) scheme in which a relay node, also equipped with AMC, retransmits redundancy packets when it is able to decode the source information packet correctly. For this scenario, we also derive the throughput and packet loss performance. Finally, we present a cross-layer AMC design approach which takes into account the hybrid ARQ protocol at the link layer. The results illustrate that including AMC in the HARQ protocols leads to a substantial throughput gain. While the performance of the AMC with HARQ protocol is strongly affected by the channel correlation, the CHARQ protocol provides noticeable performance gains over correlated fading channels as well.

Cognitive radios, which enable the coexistence on the same bandwidth of licensed primary and unlicensed secondary users, have the potential for dramatically increasing the efficiency of wireless networks. In this paper, we propose an on line learning algorithm to optimize the transmission strategy of secondary users in interference mitigation scenarios, where the secondary users are allowed to superimpose their transmission onto those of the primary users. Due to practical imitations, the secondary users have access to only a fraction of the current state of the primary users'' network. Therefore, the strategy of the secondary users is defined on a reduced state space. Numerical results show that the proposed practical learning algorithm operates close to the performance of the system under full knowledge.

This paper presents performance analysis and cross-layer design approaches for hybrid ARQ (HARQ) protocol in wireless networks which employ adaptive modulation and coding (AMC) at the physical layer and are subject to time-correlated fading channels. Utilizing a Markov channel model which accounts for the temporal correlation in successive parity transmissions by the adaptive rate HARQ protocol, we derive the system throughput and the packet loss probability based on a rate compatible punctured convolutional (RCPC) code family. As an application, we then present a cross-layer AMC design which takes into account the performance gain of the HARQ protocol at the link layer. The results illustrate that including AMC in the HARQ protocol leads to a substantial throughput gain, but the channel correlation strongly diminishes the performance of the HARQ protocol in terms of throughput and packet loss rate.

A novel framework for the analysis and optimization of cognitive wireless networks with unslotted time operations and reactive primary users is proposed. In the considered network setting, primary users' channel access is regulated by a carrier sense-based contention mechanism. As the sensing mechanism cannot distinguish between primary and secondary signals, secondary users' activity may interfere with primary users' channel contention, thus biasing the statistics of the stochastic process modeling primary users' transmissions. In fact, a primary user which wakes up during a transmission from a secondary user may sense a busy channel and enter backoff or generate a collision. The proposed framework considers these effects and optimizes the fraction of time a secondary user is allowed to transmit according to a constraint on the minimum throughput achieved by the primary users. Numerical results are presented which illustrate fundamental behaviors and tradeoffs in a network with one primary and one secondary user. Extension to more general scenarios is also discussed.

We present a framework, based on Markov models, for the analysis of error control techniques in video transmission over wireless channels. We focus on retransmission-based techniques, which require a feedback channel but also enable to perform adaptive error control. Traditional studies of these methodologies usually consider a uniform stream of data packets. Instead, video transmission poses the non-trivial challenge that the packets have different sizes, and, even more importantly, are incrementally encoded; thus, a carefully tailored model is required. We therefore proceed on two different sides. First, we consider a low-level description of the system, where two main inputs are combined, namely, a video packet generation process and a wireless channel model, both described by Markov Chains with a tunable number of states. Secondly, from a highlevel perspective, we represent the whole system evolution with another Markov Chain describing the error control process, which can feed the packet generation process back with retransmissions. The framework is able to evaluate hybrid automatic repeat request with selective retransmission, but can also be adapted to study pure automatic repeat request or forward error correction schemes. In this way, we are able to comparatively evaluate different solutions for video transmission, as well as to quantitatively assess their performance trends in a variety of scenarios. Thus, our framework can be used as an effective tool to understand the behavior of error control techniques applied to video transmission over wireless, and eventually identify design guidelines for such systems.

This paper presents an analytical framework for the study of hybrid ARQ techniques aimed at the transmission of multimedia content with differential encoding. We propose a Markov model of a selective repeat hybrid ARQ transmission scheme, where we assume that packets have different properties in terms of size, information content, and retransmission limit, so as to capture the differential encoding which characterizes such multimedia data. We consider a non-zero round-trip time channel modeled through a discrete-time Markov chain. We provide an analytical tool for the evaluation of two main performance metrics, namely, throughput and goodput. The model is extremely flexible and allows evaluation of several channel conditions and comparison of different types of ARQ (plain ARQ, Type I and II hybrid ARQ). Our results can be used as an effective tool to define design guidelines of multimedia transmission systems and to understand their performance trends.

We present a novel optimization framework based on stochastic control and Markov theory for wireless networks where users concurrently access the channel and implement retransmission-based error control. In order to let users transmit at the same time, we consider an interference mitigation, rather than a collision avoidance approach. Our focus is on the interaction between the stochastic processes modeling the various individual sources of the network. Due to retransmissions, transmission by a user does not only instantaneously interfere with other simultaneous communications, but also biases the future evolution of the stochastic processes describing the other users. We, thus, define a novel interference measure called process distortion, that takes this effect into account. We investigate the optimization of access and power control for a network with two groups of users where transmission by the second group is constrained by the process distortion generated to the first group. We present algorithms to solve the unconstrained and constrained infinite horizon average cost per stage problems modeling this scenario. We discuss in detail the application of this framework to cognitive networks.

In this paper, we investigate the effectiveness of cooperative techniques in non-infrastructured ad hoc networks based on carrier sense access control. We point out, via analysis and detailed network simulations, that CSMA biases some fundamental statistics of geographical positioning of terminals available for cooperation. Not only may relay nodes refrain from cooperating due to the carrier sense mechanism, but also the spatial correlation of interference in the network is likely to compress potential cooperators within an area in the proximity of the source. Furthermore, our discussion highlights how issues that affect carrier sense based medium access strategies, such as hidden terminals, reduce the effectiveness of relaying strategies when implemented in large non-centralized networks. The combination of the aforementioned factors is shown to significantly diminish the performance gain granted by cooperative schemes widely studied in the literature from a theoretical perspective.

The computation of the intersection area of a large number of circles with known centers and radii is a challenging geometric problem. Nevertheless, its resolution finds several applications in the analysis and modeling of wireless networks. Prior literature discusses up to three circles and even in this case there are many possible geometric configurations, each leading to a different involved close-form expression for the intersection area. In this paper, we derive two novel geometric results, that allow the check of the existence and the computation of the area of the intersection regions generated by more than three circles by simple algebraic manipulations of the intersection areas of a smaller number of circles. Based on these results, we construct an iterative algorithm based on a trellis structure that efficiently computes the intersection areas of an arbitrary number of circles. As an example of practical application of our results, we derive the probability that a fixed number of mobiles can be reliably allocated to a set of base stations in code division multiple access-based cellular networks.

In this paper, we present a methodology to obtain a channel description tailored on performance evaluation for incremental redundancy hybrid automatic repeat request schemes. Such techniques counteract channel errors by using data coding and transmitting parts of the codeword over different channel realizations. We focus on coding performance models where the error probability is asymptotically zero if the channel parameters of these realizations fall within a given region. To map this region in a compact but still precise manner, we adopt a finite-state channel model. This approach is quite common in the literature; however, differently from existing work, we propose a novel method to derive efficient channel partitioning rules, i.e., a code-matched quantization of the channel state. Such a representation enables the use of accurate Markov models to study the system performance. Compared to existing channel representation methods, our proposed technique leads to a more accurate evaluation of higher layer statistics while at the same time keeping the computational complexity low.

We consider a network with an arbitrary number of nodes, where transmission of data packet is enhanced by a medium access control protocol that allows cooperation among nodes and solves transmission failures by an automatic repeat request (ARQ) protocol. Nodes operate in half-duplex mode and may interfere with each other as no coordination is assumed. In this scenario, we first derive the outage probability for a given number of interfering nodes and then relate the interference level to the single node behavior. We analyze the network behavior with a steady state analysis that matches the number of interfering nodes with the number of ARQ retransmissions. Two alternatives for cooperation are considered: decode and forward (DF) cooperation and multiple input-single output (MISO) cooperation, where for each packet transmission either one node transmits at a given time or two cooperating nodes transmit simultaneously. In our steady state analysis we also include an opportunistic version of DF, where cooperation is activated only when the quality of the cooperator-destination link is better than that of the source-destination link. Moreover, we investigate cooperator selection based on its distance to the destination.

In this paper, we present a novel analytical framework for the study of a multiple access scheme for an ad hoc code division multiple access (CDMA) network with hybrid automatic repeat request (HARQ) error control and matched filter linear successive interference cancellation receivers. Unlike most prior work on Multi-User Detection (MUD), we directly address networking issues. In particular, we develop an approach based on renewal and semi-Markov processes, that accurately accounts for the statistical dependencies between interference, coding rate, and performance in a multiple access system. We show that our approximate approach can accurately predict throughput and failure rate values, and present results for a specific system, providing useful insight on issues related to the use of HARQ error control and MUD receivers in CDMA ad hoc networks.

This paper is meant to discuss the practical application of cooperative techniques in wireless UMTS cellular networks. In recent years cooperation in wireless networks has been the focus of a considerable research effort and several techniques have been proposed and investigated. Our feeling is that the many theoretical studies existing in the literature should now be applied to more realistic environments to pave the way to practical applications. In this paper we consider well-known cooperative techniques, such as decode & forward and amplify & forward, and we investigate and discuss the benefits their use can grant to UMTS cellular networks. In particular, we evaluate the performance of these two schemes in terms of coverage range extension and we discuss how cooperative nodes characterization influences their effectiveness.

This paper presents an analytical model for the study of hybrid ARQ techniques on discrete time Markov channels by means of an appropriate Markov chain, which tracks the transmission outcome and can be used to evaluate several performance metrics, including throughput, loss probability, number of retransmissions, and delay. The analysis is carried out with the assumptions that the information frame is encoded by the source with a linear block code and hard decoding is used at the receiver side. We finally present numerical evaluations for the performance of a truncated type II hybrid ARQ technique based on Reed Solomon erasure codes.

For a network where each node has multiple antennas, we propose a transmission mode and a cooperation protocol, with the aim of maximizing the network throughput. The distinctive feature of the work is that the focus in both the design and the evaluation is at the network level, rather than on a single link. To this end, we propose the use of spatial multiplexing and code division multiple access (CDMA) to increase the parallelism of transmissions in the network, thus improving throughput. Cooperation is also implemented by spatial multiplexing and CDMA, together with an adaptive hybrid automatic repeat request mechanism that adapts the retransmissions to the actual channel conditions. Spatial multiplexing allows frame-asynchronous transmissions and a flexible cooperation protocol that minimizes the signaling overhead. The resulting scheme is named layered coded cooperative system (LCCS). We propose an implementation of LCCS based on linear erasure packet codes, where cooperation is transparent to the receiver, and we assess the performance of LCCS both by analyzing a simple network with three nodes and by simulating a more complex network.

In this paper, we present a novel MAC approach for ad hoc networks in which Multiple Input-Multiple Output (MIMO) techniques are used at the physical level (PHY). The use of MIMO in PHY point-to-point as well as multiuser communications has been extensively studied in the recent literature. Here we go one step further, extending the approach to also include protocol design for decentralized and infrastructureless ad hoc networks. First, we study the impact of MIMO on packet transmission in an ad hoc network setting. Then, following a crossiquestlayer design paradigm, we deploy a distributed access control protocol and characterize its performance, with the aim to improve spatial reuse and convey more information on a single ad hoc link. We also explore the interaction between the Medium Access Control (MAC) and PHY layers, and use this knowledge to implement proper policies for distributed traffic control and robustness against interference. Important tradeoffs that arise when managing radio resources are highlighted, and extensive simulation results are presented and discussed.

Cooperation has proved to be an effective technique for improving the performance and the efficiency of wireless networks. Most of the existing work on cooperation focuses on the physical layer, with the aim of enhancing the capacity and the quality of a single link. In this paper we propose a cooperative protocol that melds physical, MAC and routing layers to increase the performance of a MIMO-BLAST ad hoc network, where nodes are allowed to transmit simultaneously. Nodes try to resolve in-range delivery of packets with an adaptive distributed Hybrid ARQ scheme to counteract interference from simultaneously active communications, while dynamic route selection is implemented for avoiding transmissions over links with harsh fading conditions. We assess the performance of our scheme through detailed simulations.

In this paper we design a network based on cooperative spatial multiplexing (SM), capable to adaptively support terminals with multiple antennas and multiple cooperating nodes. Our proposal overcomes limitations of existing cooperative networks based on space-time block codes (STBC), i.e., their need for symbol synchronization and signalling overhead for cooperation setup. Furthermore, the presented cooperation scheme is designed with the aim of improving the overall network performance, while most of the proposals in the literature focus on single link performance. SM is integrated with decision- feedback multiuser receivers for both the non-cooperative and the cooperative phases in order to maximize the use of radio resources. The cooperation efficiency is further enhanced by using a hybrid automatic repeat request (ARQ) mechanism for error control, implemented with linear erasure packet codes. Also in this case, the packet-based coding allows to have no signalling overhead for cooperation. Extensive results are provided, obtained by simulation of the complete PHY and MAC layers for a realistic network scenario with several nodes.

We derive the expressions of the outage probability for a wireless network that integrates hybrid automatic repeat request (HARQ) and coded cooperation among nodes. The medium access control (MAC) provides a HARQ protocol where the source node, upon a decoding failure at the destination node, transmits additional coded bits for the same data packet. If a neighbor node is able to decode the first transmission, it cooperates with the source by sending additional coded bits. In order to keep nodes simple, we assume half-duplex, single band asynchronous transmissions, and receivers with single user detection techniques. Various configurations of cooperation are considered and the analysis includes key characteristics of wireless communications and interference.

In this paper, we consider a MAC protocol for ad hoc networks where nodes are equipped with multiple antennas, and communications are spatially multiplexed using the Bell labs LAyered space time (BLAST) system. The contribution of this paper is twofold. First, we introduce two different analytical models aimed at predicting the propagation of detection errors within the BLAST receiver, the first based on a Gaussian approximation of the detection errors, and the second on the weighed enumeration of error configurations. A simplification of the latter, with lower complexity, is also obtained and compared to the original model. We then use these analytical tools to assess the performance of a cross–layer MAC protocol, and compare it with fully detailed simulations. Since the analytical tools replace the simulation of the physical layer, the proposed semianalytical approach is much faster than the bit-by-bit simulations. Numerous results are provided for the network performance assessment, showing that our semianalytical approach is able to predict network behavior with very good accuracy, but much lower complexity.

This paper presents an analytical model for the analysis of hybrid ARQ techniques on discrete time Markov channels by means of Markov chains. The first contribution is an original proposal to track the outcome of the packet transmissions in a generalized hybrid ARQ transmission system, deriving an appropriate Markov chain. Also, an example is given of how to put this Markov chain in relationship with a discrete-time Markov channel description. This framework can be used to evaluate from the theoretical point of view the performance of truncated type II hybrid ARQ techniques, deriving in general some useful insight on the behavior of such systems.

In this paper, we address the impact of different access strategies in ad hoc networks with multiple antennas and MIMO communications. We employ a cross-layer designed MAC protocol that allows both for multiple simultaneous access to the radio medium and for proper exploitation of multiuser detection at the receiver for interference cancellation purposes. Still, the network is subject to early deadlock if no access strategy is employed that reduces transmission persistency. To this aim, we study two types of exponential backoff, namely node- wise and destination-wise, and combine them with a form of cooperative agreement on who takes the role of transmitter or receiver. The impact of all schemes is assessed and a comparison is pursued, focusing on typical network metrics (such as throughput and transmission delay among others) and showing when and why one technique performs better than the others.

In this paper, we consider a layered space-time multiuser detection technique and propose an analytical approximation for its performance. Our work is useful in two different stages of network design. On one hand, the approximation may be used to evaluate the bit and packet error performance of communications among a group of terminals making use of multiuser detection. On the other hand, it can also be seen as a very valuable tool from the networking point of view, as it may help in designing radio access control protocols based on multiuser detection, as analytical formulas are very fast to evaluate, in contrast to bit-level simulations that may need a long time to complete. Thus, an analytical formulation is important, because it helps discriminating among different protocol alternatives, speeding up considerably the protocol design phase and the development of new radio access policies for multiuser networks.

In this work, we analyze the effects of a distributed transmission coordination scheme that is particularly suited for ad hoc networks with the ability to exploit spatially--multiplexed communications over MIMO links. Following previous work where we discussed the performance of this kind of networks and deployed a fast yet reliable approximation for physical layer behavior, we now employ this knowledge to analyze the performance of a different access scheme, meant to outperform previous results by the use of distributed coordination of transmissions and receptions without adding any further redundancy in communications. Furthermore, we analyze the effect of tuning two parameters on the overall network performance.

We consider the application of spatial multiplexing to ad hoc networks where nodes have multiple antennas. At the physical level, we suppose that layered space-time multiuser detection (LAST-MUD) is applied to separate multiple streams arriving at the receiver simultaneously. Our contributions here consist first in the reproduction of the multiuser detection process performance by a simple analysis, where we also specify when the analytical results are expected to be accurate and why. Second, we use this approximation to perform a cross-layer design of a MIMO ad hoc network where physical layer and medium access control strategies are integrated to maximize the network throughput. We finally corroborate our conclusions, by comparing analysis with simulation results both at the link and the network level

Existing cooperative diversity techniques for wireless ad hoc networks mostly consider space-time block codes for cooperation. In this paper we propose a cross-layer design of ad hoc wireless networks based on spatial multiplexing (SM). Each node is equipped with multiple antennas and the spatial dimensions of the channel are exploited to support multiple simultaneous transmissions. Diversity is then provided only for failed transmissions by means of selective cooperation among nodes that still use SM for transmission. Moreover, to increase the efficiency of the system, a hybrid automatic repeat request (HARQ) protocol integrated with packet coding is used at the medium access control layer. We analyze the proposed network architecture with a Markov chain description of the decoding process and we derive a closed form expression for the achieved throughput in Rayleigh fading channels.

In this paper, we shed some light on the implications that using a recently proposed layered space-time multiuser detection technique has on MAC protocol design for ad hoc networks with multiple antennas. From this point of view, our work relates to both physical layer and network layer studies. In fact, on the one hand physical layer aspects are important to characterize the behavior at the receiver, especially in terms of bit and packet error rates, but are rarely considered in conjunction with networking scenarios; on the other hand, networking aspects are typically studied using drastically simplified physical layer models that, while allowing easier networking analysis, are often too restrictive or unrealistic. Also, this disconnects between physical and networking layer studies may severely limit the possibilities for cross-layer optimization, which appears to be the right approach for efficient wireless network design. Our paper is then an effort to establish some connection between the "physical" and "network" approaches, highlighting some interesting capabilities that ad hoc networks with multiple antennas are endowed by the use of multiuser detection

In this paper, we propose a novel technique for the provisioning of multicast streaming flows in 3G W-CDMA cellular systems. Our focus here is on the transmission of a downlink multicast streaming flow to the interested users in a 3G cell. For error control, we propose packet-based forward error correction (FEC) and with the transmission of a certain amount of redundancy over parallel common channels. In practice, we exploit the temporal diversity over multiple channels to strengthen the FEC scheme thereby increasing the QoS. To increase error resilience, in every channel we exploit well-known packet-based Reed Solomon like coding techniques. Moreover, appropriate time shifts of the information sent across the parallel common channels are also introduced to increase the robustness against error bursts. The performance evaluation is carried out through an analytical framework. The obtained results show the substantial benefits deriving from the usage of a multiple channel transmission technique.