This paper presents the preliminary results of a Linear Parameter Varying-Autoregressive eXogenous model, identified through Least Squares Support Vector Machines, able to optimally drive a robotic arm system by emulating the performance of a Nonlinear Model Predictive Control (NMPC) policy. The support vector machine framework is employed to replicate the control performance of a computationally demanding NMPC. Due to the nonlinear characteristics of the robotic arm, the NMPC is suitable to guarantee expected control performance. However, its application in real-time systems with fast dynamics is limited by high memory and computational demands required at each sampling instant. In this work, the linear parameter varying model is trained using a data-driven approach to imitate the control actions of the NMPC across different scenarios. The proposed controller and the original NMPC are evaluated in simulation, considering multiple operating conditions of the robotic arm. The control performance of both approaches is then compared to assess the effectiveness of the proposed method.

Lithium-ion batteries represent a pivotal component within contemporary energy storage solutions, exhibiting a diverse range of applications spanning from consumer electronics to electric vehicles and renewable energy systems. Nevertheless, the progressive degradation of these batteries, resulting in a reduction in capacity and performance, poses significant challenges in terms of system safety and reliability. In this context, the evaluation of the Remaining Useful Life (RUL) plays a central role in assessing the health of lithium-ion batteries. Ensuring precise and reliable RUL prediction is critical for the proper operation of a system. In this work, to address these challenges, a novel lightweight deep learning approach has been proposed for battery RUL estimation, by using voltage and current data. The proposed model is an approach based on the Echo State Networks (ESNs), which is compared to conventional deep learning models, such as Long Short-Term Memory (LSTM) networks, which re-quire more complex architectures and substantial computational resources. The ESN-based model demonstrates a comparable predictive capacity, while substantially reducing training and inference times. The model was tested with the CALCE dataset, focused on data obtained during charge and discharge cycles of lithium-ions batteries. Specifically, under the test conditions of 1 C discharge, the ESN requires only 0.3 seconds for training and approximately 0.06 seconds for inference, thus offering a computational advantage over the LSTM model, which requires 384 seconds for training and approximately 0.19 seconds for inference with the same hardware.

Advances in machine learning (ML) have opened up new opportunities to include decision-making capabilities in Internet of Things (IoT) nodes. This opportunity is complex to address, since conventional ML implementations are computationally intensive, thus reducing the possibility of implementing them on resource-constrained systems. In addition, when nodes in a network grow significantly, centralized data processing creates new challenges such as latency, resource efficiency, privacy, bandwidth, etc. Aiming to simultaneously address the above challenges, this paper presents a novel efficient distributed learning strategy for ESN-Based model, that is conceived for training and inference to incorporate part of the model while reducing the sharing of private information. The experiment setup was conducted using Appliances Energy Prediction dataset to forecast energy consumption under different environmental and usage conditions, using simulated IoT devices. The numerical results show that the proposed solution performs well compared to a classical centralized ESN-based approach without sacrificing too much performance at very low computational costs.

Human-robot interaction and affective computing are cross-disciplinary fields whose connections are being further explored with the proliferation of humanoid robots that assist people in their daily activities. The proposed study aims to assess emotion recognition accuracy based on facial image acquisition, recorded by the camera sensor of the humanoid robot Pepper, during a human-robot interaction scenario. The emotions of the individuals involved were elicited by viewing five different video clips, each corresponding to different target emotions (e.g., happiness, sadness, anger, surprise, and neutrality). Emotion evaluation was carried out using the Pepper integrated classification module and convolutional neural networks, and the obtained results were compared to determine the best emotion classification accuracy. The test group was diverse, particularly in terms of participant age, including a significant number of elderly individuals who were not familiar with robot interaction. This diversity also allowed for an examination of user acceptability in the context of the human-robot interaction experience.

Machines are increasingly getting incorporated into real-life scenarios involving humans. From small devices offering basic aids to big humanoid robots designed to assist humans, involvement is on the rise. Increased integration and consequent interaction between robots and humans are the future directions we are headed in. In this paper, we explore the Pepper humanoid robot as an assistive robot in a scenario where support is needed in the form of a specific task and the robot assistant moves autonomously. We present a description of the design and development of the system, the findings, and the inferred future directions for research and development in the domain. An assistive robot system that meets requirements such as independence, good perception, navigation, communicative ability, and adaption to changing ambiance could be extended to various environments. Implementing a use case of monitoring a precautionary measure of dressing in a face mask, we explore Pepper as an assistive robot effective for such tasks.

This paper presents a preliminary study considering the design of an adaptive torque allocation policy for electric vehicles combining optimal control with data-driven techniques. The vehicle is equipped with four independent actuated wheels driven by electric motors. The policy aims to control the vehicle powertrain by allocating available power among motors to satisfy the driver control torque request and adjust torque allocated to different motors according to estimated wheels slip ratio change due to terrain varying conditions. A constrained optimal torque allocation algorithm is designed to distribute available power among wheels. In order to adjust the power allocation result, a data-driven adaptive policy is designed to adjust the control allocation parameters and the torque distribution reflecting wheel's operating conditions. The combination of torque allocation and data-driven adaptation policies permits the adjustment of the allocated power according to the wheel/road contact conditions. The algorithm has been tested and validated in simulation, showing the improvement given by the proposed approach compared with respect to the control system neglecting the data-driven adaptation of the torque allocation policy.

This article presents a Fault Detection (FD) method to deal with propeller faults on multirotor drones in real-time. Several solutions have been proposed in the literature, however, they depend on additional sensors and/or dedicated hardware to deal with heavy computational complexity. So, they cannot be implemented in off-the-shelf commercial devices, i.e., without the aid of additional on-board sensors and/or extra computational power. The proposed method, instead, requires the on-board Inertial Measurement Unit (IMU) data only: by combining Finite Impulse Response (FIR), together with sparse classifiers, only a subset of the features is actually needed online and the FD is thus feasible in real-time. Design and tests are based on real flight data from a hexarotor, equipped with a conventional ArduPilot-based controller. The classification accuracy in testing is up to 93.37% (98.21%) with a binary tree (Linear Support Vector Machine (LSVM)). Moreover, the space and time complexity of the proposed method is low: on a PixHawk Cube flight controller, it requires less than 2% of the cycle time, and can then run in real-time. Finally, the proposed fault detection solution is model-free and it can be easily generalized to other multirotor vehicles.

The present paper proposes an architecture for a product process monitoring system suitable for SMEs (Small-Medium Enterprises). The monitoring system is the main means by which decision-making systems based on intelligent automation technologies are aware of the state of the system on which they will take decisions. Methods and tools from best-practice and best-effort approaches are proposed in the context of SMEs, where the requirements of low cost, low initial level of digitalization and high production flexibility often coexist and contribute to the complexity of management and control problems in these companies. The paper focuses on the design of the monitoring system using an OSGi framework to meet industry standards and Industry 4.0 requirements, taking into account the peculiarities of SMEs as design constraints. The proposed architecture was first tested using a simulation tool and then implemented on a full-scale production line used for data collection.

This article presents a fault diagnosis algorithm for rotating machinery based on the Wasserstein distance. Recently, the Wasserstein distance has been proposed as a new research direction to find better distribution mapping when compared with other popular statistical distances and divergences. In this work, first, frequency- and time-based features are extracted by vibration signals, and second, the Wasserstein distance is considered for the learning phase to discriminate the different machine operating conditions. Specifically, the 1-D Wasserstein distance is considered due to its low computational burden because it can be evaluated directly by the order statistics of the extracted features. Furthermore, a distance weighting stage based on neighborhood component features selection (NCFS) is exploited to achieve robust fault diagnosis at low signal-to-noise ratio (SNR) conditions and with high-dimensional features. In detail, the NCFS framework is here adapted to weight 1-D Wasserstein distances evaluated from time/frequency features. Experiments are conducted on two benchmark data sets to verify the effectiveness of the proposed fault diagnosis method at different SNR conditions. The comparison with state-of-the-art fault diagnosis algorithms shows promising results.Note to Practitioners—This article was motivated by the problem of fault diagnosis of rotating machinery under low SNR and different machine operating conditions. The algorithm employs a statistical distance-based fault diagnosis technique, which permits to obtain an estimation of the fault signature without the need for training a classifier. The algorithm is computationally efficient during the training and testing stages, and thus, it can be used in embedded hardware. Finally, the proposed methodology can be applied to other application domains such as system monitoring and prognostics, which can help to schedule the maintenance of rotating machinery.

Recurrence quantification analysis (RQA) allows to quantify the periodic behavior using recurrence plots instead of deriving information purely from visual analysis. The current study presents a preliminary analysis of stator-current measurements for electric motor fault detection and classification by means of the recurrence quantification theory. Firstly, a preliminary visual inspection of the recurrence plots of stator-current measurements for healthy and faulty electric motors is presented. Thereafter, the following RQ metrics are analyzed: the recurrence rate, the determinism, the divergence, the Shannon entropy, the laminarity and the trapping time. Then, the RQ metrics are used as predictors for fault detection and classification. The classification results (100% fault classification accuracy), which are presented using the linear support vector machine classifier, show that the RQA can be considered as a tool for motor current signature analysis.

Nowadays the inspection of products at the end of line represents a critical phase. At this stage, it is necessary to look for defects in order to prevent the quality check to fail, and to provide information for improving the production as well. This task can be performed by using several sensing technologies, and the contactless optical inspection plays a key role. In this regard, the use of advanced robotic manipulators offers the capability to change the viewpoint of a given object and to inspect its multiple faces. We propose an approach that combines the use of photometric stereo to derive a 3D model of objects, empowered by the super-resolution that is applied on the original dataset (upstream) or on the normal images (downstream) in order to increase the quality of the final 3D model. The vision system is mounted on a robotic manipulator, able to grasp and change the viewpoint, thus offering a more complete view of the object to be inspected. The obtained results show that the developed solution increases the quality of the derived 3D models used for inspection tasks on different faces of the objects; this is achieved by using the manipulation ability offered by the adopted robotic platform.

Here, the authors present a low cost telerehabilitation system made up of a commercial red–green–blue depth (RGB‐D) camera and a web‐based platform. The authors goal is to monitor and assess subject movement providing acceptable and usable at‐home remote rehabilitation services without the presence of a clinician. Clinical goals, defined by physiotherapists, are firstly translated into motion analysis features. A Takagi Sugeno fuzzy inference system (FIS) is then proposed to evaluate and combine these features into scores. In this stage, the ‘collaborative design’ paradigm is used in depth and complete manner: the contribution of the clinician is not limited only to the rules definition but enters in the core of the evaluation algorithm through the definition of the fuzzy rules. A case study on low back pain rehabilitation involving 40 subjects, 5 exercises, and 4 physiotherapists is then presented to the effectiveness of the proposed system. Results of the validation of the system aimed at the assessment of the reliability of the proposed approach show high correlations between clinician evaluation and FIS scores. In this scenario, due to the high correlation, each FIS could represent a virtual alter‐ego of the physiotherapist which enable a real time and free second opinion.

This paper proposes a control methodology based on Dynamic Surface Control (DSC) to manage the power flow of a photovoltaic (PV) battery system. In particular, due to the inner stochastic nature and intermittency of the solar production and in order to face the irradiance rapid changes, a robust and fast controller is needed. Dynamic Surface Control is a modified version of Backstepping control that avoids the explosion of terms, which is a typical drawback of the Backstepping control and furthermore it is not affected by the well known problem of chattering, which affects Sliding Mode controllers. Dynamic Surface Control is compared to the conventional Proportional-Integral-Derivative controller (PID). In particular, DSC shows better performances in terms of steady state chattering and transient response, as confirmed by the Integral of the Absolute value of Error (IAE), Integral of the Squared Error (ISE) and Integral of Time multiplied by the Absolute value of Error (ITAE) performance indexes.

In recent years several statistical methods have been applied to condition monitoring of various processes under linearity and stationarity assumptions. However most of the actual industrial processes, e.g. in the chemical sector, are strongly nonlinear. Furthermore the hypothesis of data Gaussian distribution does not often hold, thus causing a decrease of the fault detection accuracy. In this paper a Switching Linear Regression (SLR) approach is firstly proposed in a fault detection scenario. The basic idea is to estimate different Linear Regression models through an arbitrary clustering algorithm and then switching among these models. The developed algorithm allows to deal with nonlinear processes. The proposed fault detection approach is applied to two simulated test bench and on the Tennessee Eastman process benchmark. Furthermore, compared with the Linear Regression algorithm, SLR shows better performance in terms of fault detection accuracy.

Early detection of abnormalities for electrical motors is a key point to reduce economic losses caused by unscheduled maintenance and shutdown time. In this context, health monitoring and fault diagnosis are crucial tasks to be performed. We introduce a novel Linear Discriminant Analysis (LDA) based algorithm to deal with fault data dimension reduction and fault detection issues. In particular the algorithm, namely Δ-LDA, is designed to overcome the problem of a between-class scatter matrix trace very close to zero. Indeed, if the information of the expected value is not sufficient to discriminate the classes, we propose the use of the difference of covariance matrices. A performance comparison with other conventional methods, e.g. principal component analysis and classical LDA, is proposed. In particular experimental results show that the proposed algorithm improves the classification accuracy if the classes are overlapped, and gives comparable results in the remaining scenarios.

Simulators for Wireless Sensor Networks (WSNs) are one of the most important tools for systems development. They enable to study and evaluate new theories and hypotheses for sensors data gathering, testing new applications and protocols. Nowadays, there are a large number of open source WSN simulators and they can be divided into different categories according to their features and main applications. Due to the ability to increase the real WSN prototyping, the Cross Levels Simulator, like Cooja, has become an important class of simulators. Although they are open source, flexible and extensible in all levels, the test interface, the external connection at a physical level and the direct interaction with the process control via the WSN is very poor. In this work we present the Cooja Advanced Sky Interface which is an extension of the Contiki's Cooja network simulator for the Sky mote. Due to the absence of the analog output control in the Contiki OS for the Sky mote, as additional contribution, the Contiki Sky DAC driver has been developed and tested in the Cooja Simulator with the Advanced Sky GUI and GISOO plugin to give the ability to implement control over the wireless sensor network.

This paper presents theoretical details of a model-based sensor fault detection and isolation system (SFDIS) applied to unmanned ground vehicles (UGVs). Structural analysis is applied to the nonlinear model of the vehicle for residual generation. Two different solutions have been proposed for developing the residual evaluation module. The vehicle sensor suite includes a global positioning system (GPS) antenna, an inertial measurement unit (IMU), and two incremental optical encoders.

This paper presents implementation details of a model-based sensor fault detection and isolation system (SFDIS) applied to unmanned ground vehicles (UGVs). Structural analysis, applied to the nonlinear model of the UGV, is followed to build the residual generation module, followed by a residual evaluation module capable of detecting single and multiple sensor faults, as detailed in part I (Monteriu et al., 2007). The overall proposed sensor fault detection and isolation system has been tested in real-time on the ATRV-Jr mobile robot when following different trajectories in an outdoors environment. The robot sensor suite includes a global positioning system (GPS) antenna, an inertial measurement unit (IMU), and two incremental optical encoders