After reviewing the current state of explainable Artificial Intelligence (XAI) capabilities in artificial Intelligence (AI) systems developed for critical domains like criminology, engineering, governance, health, law and psychology, this paper proposes a domain-independent Accountable explainable Artificial Intelligence (AXAI) capability framework. The proposed AXAI framework extends the XAI capability to let AI systems share their decisions and adequately explain the underlying reasoning processes. The idea is to help AI system developers overcome algorithmic biases and system limitations through incorporation of domain independent AXAI capabilities. Moreover, existing XAI methods would neither separate nor quantify measures of comprehensibility, accuracy and accountability so incorporating and assessing XAI capabilities remains difficult. Assessment of the AXAI capabilities of two AI systems in this paper demonstrates that the proposed AXAI framework facilitates separation and measurement of comprehensibility, predictive accuracy and accountability. The AXAI framework allows for the delineation of AI systems in a three-dimensional AXAI space. It measures comprehensibility as the readiness of a human to apply the acquired knowledge. The system accuracy is measured as the ratio of the test and training data, training data size and the observed number of false-positive inferences. Finally, the AXAI framework measures accountability in terms of the inspect ability of the input cues, the processed data and the output information, for addressing any legal and ethical issues.

The road to extreme low latency urges the invention of intelligent access schemes that depart from the shortcomings associated with the classic ones. In this work, we present a novel traffic prediction and fast uplink framework for IoT devices activated by binary Markovian events. First, we apply the forward algorithm in the context of hidden Markov models (HMM) to predict the hidden states, which are used to schedule the available resources to the devices with maximum likelihood activation probabilities via fast uplink grant. In addition, we evaluate the regret metric as the number of missed resource allocation opportunities to evaluate the performance of the prediction. Next, we formulate a fairness optimization problem to minimize the age of information while keeping the regret as minimum as possible. Finally, we propose an iterative algorithm to estimate the model hyperparameters (activation probabilities) in a real-time application and apply an online-learning version of the proposed traffic prediction scheme. Simulation results show that the proposed algorithms outperform baseline models such as time division multiple access (TDMA) and grant-free (GF) random-access in terms of regret, the efficiency of system usage, and age of information.

Abstract—Bulk data transfer applications are greedy in nature in that they have virtually unlimited bandwidth requirements. If serviced by a best effort congestion control, they can create periodic packet loss, long delay and throughput degradation of cross traffic, which may disrupt the normal operation of concurrent QoS sensitive applications. A Lower than Best Effort (LBE) congestion control can make greedy applications only utilize bandwidth left by QoS sensitive applications, thus improve QoS. We propose FlexiS – a Flexible Sender side LBE congestion control mechanism. Unlike most existing LBE congestion controls, which use base delay in incipient congestion detection, FlexiS employs a trend of delay based congestion detector that is more robust to dynamic network environments, in which base delay is difficult to estimate. A non-linear increase and multiplicative decrease controller is used to adapt sending rate by FlexiS. Extensive emulation and Internet tests showed that: (1) FlexiS generally has much lower impact on coexisting best effort traffic than LEDBAT; (2) FlexiS is scalable across a wide range of available bandwidth; (3) aggregate FlexiS traffic can effectively utilize available bandwidth; (4) in most situations, FlexiS flows with the same or different RTTs can fairly share available bandwidth – achieving a fairness index greater than 0.99; (5) the performance of FlexiS is almost not affected by route change and active queue management.

We consider future cellular networks and study the coexistence of delay-sensitive (DS) and delay-tolerant (DT) devices in machine-type communication (MTC). DS devices require to minimize access delay for their low delay requirements; in contrast, DT devices have flexible delay constraints.

The paper is under final review on IEEE Transaction on Network and Service Management. Innovative traffic engineering functions and services require disrupting routing and forwarding mechanisms to be performed with low overhead over complex network topologies. Source routing (SR) is a prominent alternative to table-based routing for providing the needed expressiveness and agility by reducing the number of network states. This work proposes the M-PolKA, a topology-agnostic multipath source routing scheme and orchestration architecture for reliable communications, which explores special properties from the Residue Number System (RNS) polynomial arithmetic. A P4-based proof-of-concept is experimentally demonstrated using emulated and hardware prototypes. Also, use cases for revealing M-PolKA’s functionalities are tested in different scenarios in order to address problems, such as communication reliability improvement, agile path migration and fast failure reaction. Finally, low overhead for extra functionalities is observed when RNS-based SR is compared to traditional routing approaches.

Unmanned Aerial Vehicles (UAVs) have emerged as suitable platforms for transporting and positioning communications nodes on demand, including Wi-Fi Access Points and cellular Base Stations. This paved the way for the deployment of flying networks capable of temporarily providing wireless connectivity and reinforcing coverage and capacity of existing networks. Several solutions have been proposed for the positioning of UAVs acting as Flying Access Points (FAPs). Yet, the positioning of Flying Communications Relays (FCRs) in charge of forwarding the traffic to/from the Internet has not received equal attention. In addition, state of the art works are focused on optimizing both the flying network performance and the energy-efficiency from the communications point of view, leaving aside a relevant component: the energy spent for the UAV propulsion. We propose the Energy and Performance Aware relay Positioning (EPAP) algorithm. EPAP defines target performance-aware Signal-to-Noise Ratio (SNR) values for the wireless links established between the FCR UAV and the FAPs and, based on that, computes the trajectory to be completed by the FCR UAV so that the energy spent for the UAV propulsion is minimized. EPAP was evaluated in terms of both the flying network performance and the FCR UAV endurance, considering multiple networking scenarios. Simulation results show gains up to 25% in the FCR UAV endurance, while not compromising the Quality of Service offered by the flying network.

StyleGAN is now capable of achieving excellent results, especially high-quality face synthesis. Recently, some studies have tried to invert real face images into style latent space via StyleGAN. However, morphing real faces via latent representation is still in its infancy. Training costs are high and/or require huge samples with labels. By adding regularization to style optimization, we propose a novel method to morph real faces based on StyleGAN. To do the supervised task, we label latent vectors via synthesized faces and release the label set; then we utilise logistic regression to fast discover interpretable directions in latent space. Appropriate regularization helps us to optimize both latent vectors (faces and directions). Moreover, we use learned directions under different layer representations to handle real face morphing. Compared to the existing methods, our method faster yields a larger diverse and realistic output. Code and cases are available at \url{https://github.com/disanda/RFM}.

A multizone sound field control method, called amplitude matching, is proposed. The objective of amplitude matching is to synthesize a desired amplitude (or magnitude) distribution over a target region with multiple loudspeakers, whereas the phase distribution is arbitrary. Most of the current multizone sound field control methods are intended to synthesize a specific sound field including phase or to control acoustic potential energy inside the target region. In amplitude matching, a specific desired amplitude distribution can be set, ignoring sound propagation directions. Although the optimization problem of amplitude matching does not have a closed-form solution, our proposed algorithm based on the alternating direction method of multipliers (ADMM) allows us to accurately and efficiently synthesize the desired amplitude distribution. We also introduce the differential-norm penalty for a time-domain filter design with a small filter length. The experimental results indicated that the proposed method outperforms current multizone sound field control methods in terms of accuracy of the synthesized amplitude distribution.

This paper introduces shape deformation as a new approach to the computer-aided design (CAD) of high frequency components. We show that geometry deformation opens up new design possibilities and offers additional degrees of freedom in the 3D modeling of microwave structures. Such design flexibility is highly desirable if the full potential of Additive Manufacturing (AM) is to be exploited in the fabrication of RF & microwave devices. The use of deformation techniques in the design of high-frequency components allows the attainment of improved electrical parameters, such as high-quality factors for cavity resonators and wide higher-order mode separation. In this work, shape deformation with radial basis functions (RBFs) is integrated with an electromagnetic field simulator based on the 3D finite-element method (FEM), allowing the semiautomated optimization of microwave components, such as cavity resonators and filters. The proposed strategy is used for the design of high Q-factor cavity resonators, cavity bandpass filters with improved spurious mode separation and of a compact twisted waveguide filter. Three designs of waveguide cavity filters with complex geometry are experimentally verified using 3D-printed prototypes fabricated with the selective laser melting (SLM) technology.

Radio frequency-orbital angular momentum (RF-OAM) is a novel approach of multiplexing a set of orthogonal modes on the same frequency channel to achieve high spectrum efficiencies. Since OAM requires precise alignment of the transmit and the receive antennas, the electronic beam steering approach has been proposed for the uniform circular array (UCA)-based OAM communication system to circumvent large performance degradation induced by small antenna misalignment in practical environment. However, in the case of large-angle misalignment, the OAM channel capacity can not be effectively compensated only by the electronic beam steering. To solve this problem, we propose a hybrid mechanical and electronic beam steering scheme, in which mechanical rotating devices controlled by pulse width modulation (PWM) signals as the execution unit are utilized to eliminate the large misalignment angle, while electronic beam steering is in charge of the remaining small misalignment angle caused by perturbations. Furthermore, due to the interferometry, the receive signal-to-noise ratios (SNRs) are not uniform at the elements of the receive UCA. Therefore, a rotatable UCA structure is proposed for the OAM receiver to maximize the channel capacity, in which the simulated annealing algorithm is adopted to obtain the optimal rotation angle at first, then the servo system performs mechanical rotation, at last the electronic beam steering is adjusted accordingly. Both mathematical analysis and simulation results validate that the proposed hybrid mechanical and electronic beam steering scheme can effectively eliminate the effect of diverse misalignment errors of any practical OAM channel and maximize the OAM channel capacity.

Measurements from an in-service immersed tunnel. These data are used to design a Wireless Sensor Node for Structural Health Monitoring (SHM) system.

Invasion of locust swarms has affected the crops in many countries in Africa and Asia, which is a significant threat to food security. Therefore, different approaches are adopted to monitor and control the locust swarms to save the crops. Furthermore, it has been proved in various studies that technology can help in agriculture through drones, real-time data monitoring, or teaching the farmers with the latest tools. Following the UN sustainability goals for food security, this research has presented a Virtual Reality(VR) based educational application to teach sustainable locust management strategies. Using hand tracking technology in the Oculus Quest lets users learn how farmers can deal with locusts without pesticides. Based on a storytelling approach, the methods presented are profitable for the farmers and free of any harm to crops regarding food security. This application can help motivate the adoption of these sustainable locust control strategies in broader interventions for environmental recovery.

Wireless fully-passive neural recording systems are demonstrated with heterogeneous integration of sensing, mixing and communication components in flexible polymer dielectric films. The recording neurosensor has an antenna, antiparallel diode, and a by-pass capacitor to modulate an incoming carrier signal with the neuropotentials and backscatter the mixed signal to an external reader circuit. Planar antennas are realized on Liquid Crystal Polymer (LCP) flexible substrates with low dielectric constant. The designed antenna topology is realized with a single metal layer and eliminates the dependency on substrate thickness and leads to thinner sensors. This is a major advance compared to prior passive neural recording with rigid and thicker substrates and components. Approximately 80% thickness reduction and 20% volume reduction are achieved with similar performance compared to earlier studies. Our approach thus leads to low-cost and disposable wireless neurosensors on a flexible platform with heterogenous integration.

The increased complexity and intelligence of automation systems require the development of intelligent fault diagnosis (IFD) methodologies. By relying on the concept of a suspected space, this study develops explainable data-driven IFD approaches for nonlinear dynamic systems. More in detail, we parameterize nonlinear systems through a generalized kernel representation used for system modeling and the associated fault diagnosis. An important result obtained is a unified form of kernel representations, applicable to both unsupervised and supervised learning. More importantly, through a rigorous theoretical analysis we discover the existence of \emph{a bridge} (i.e., a bijective mapping) between some supervised and unsupervised learning-based entities. Notably, the designed IFD approaches achieve the same performance by the use of this bridge. In order to have a better understanding of the results obtained, unsupervised and supervised neural networks are chosen as the learning tools to identify generalized kernel representations and design the IFD schemes; an invertible neural network is then employed to build the bridge between them. This study is a perspective article, whose contribution lies in proposing and detailing the fundamental concepts for explainable intelligent learning methods, contributing to system modeling and data-driven IFD designs for nonlinear dynamic systems.

This work has been submitted to the IEEE for possible publication. Copyright may be transfered without notice, after which this version may no longer be accessible. We introduce the Physiologically-informed Gaussian Process (PhGP) model, a novel Bayesian probabilistic approach to integrate and interpret prior physiological knowledge in machine learning models. Existing recognition algorithms often consider either end-end-to models and/or alongside feature extraction techniques. Conversely, our model based on Gaussian Processes (GP) which are grounded on Bayesian statistics proposes a principled and interpretable integration of these techniques for recognition problems in biomedical engineering realm. This paper builds upon but significantly extends our previous conference paper presented at EMBC 2020 [1]. In this paper, we develop a new model that considers both the raw physiological signals and the prior expert knowledge for training GP models. Moreover, unlike our previous paper by relying on the explicit formula of the GP inference equations, we developed an interpretability framework for our proposed recognition model. Finally, we thoroughly validated our new model on two different public datasets. Our paper brings novelties in the field of biomedical engineering, affective computing , machine learning and computer sicence. [1] Ghiasi, S., Patane, A., Greco, A., Laurenti, L., Scilingo, E.P. and Kwiatkowska, M., 2020, July. Gaussian Processes with Physiologically-Inspired Priors for Physical Arousal Recognition. In 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC) (pp. 54-57). IEEE.

Non-orthogonal multiple access (NOMA) has received an enormous attention in the recent literature due to its potential to improve the spectral efficiency of wireless networks. For several years, most of the research efforts on the performance analysis of NOMA were steered towards the ergodic sum rate and outage probability. More recently, error rate analysis of NOMA has attracted massive attention and sparked massive number of researchers whose aim was to evaluate the error rate of the various NOMA configurations and designs. Therefore, the large number of publications that appeared in a short time duration made highly challenging for the research community to identify the contribution of the different research articles. Therefore, this work aims at surveying the research work that considers NOMA error rate analysis and classifying the contributions of each work. Therefore, work redundancy and overlap can be minimized, research gabs can be identified, and future research directions can be outlined. Moreover, this work presents the principles of NOMA error rate analysis.

Measurement of vertebral spinal motion is a difficult endeavor given the lack of reachable bony landmarks for placement of reflective markers for common optoelectronic systems and large skin movement artifacts along the spine affecting the accuracy of surface-mounted techniques. The objective of this proof-of-concept study was to develop an experimental framework to assess spinal curvatures with high accuracy by using ultrasound (US) imaging. A phantom spine model (T1-T12) immersed in water was scanned using both Magnetic resonance imaging (MRI) and three dimensional (3D) US. The US scans were obtained for the phantom spine in different postures while the US probe was localized by three electromagnetic tracking sensors (NDI system) attached to the probe head and later identified on the acquired US volumes via visual inspection. Spinal landmarks were selected on the US images and validated using US-MRI registration. Their position was then related to the position of the tracking sensors, enabling the precise localization of landmarks and reconstruction of spinal curvatures in a global reference frame. The obtained anatomical reconstructions were further refined using an automatic registration algorithm and validated by an experienced sonographer (ground-truth solution). Results of this study show that localized 3D US enables tracking of spinal landmarks and allows for an accurate reconstruction of static postures. A high reliability in sensor identification of 0.49 mm was found on average while the inter-observer variability was strongly correlated to the intra-observer error with maximum deviance of 0.8 mm. The NDI-based spine reconstruction had a mean error of 5.6 mm compared to the ground-truth, which is significantly lower than errors obtained by current gold standards (>10 mm). Combining US probe tracking and semi-automatic registration methods anatomical reconstructions comparable to clinical standards were obtained. This study implies that US holds the potential for advancing vertebral motion assessment and the reconstruction of spinal curvatures. In future, this could overcome limitations of marker-based approaches.

Published in IEEE Transactions on Information Forensics and Security (T-IFS) on 23 June 2022. This version on TechRxiv contains supplemental material.