Robust control of magnetic levitation (maglev) plant remains a significant challenge due to its inherent non-linearities and uncertainty to exogenous perturbations, though maglev technology has a wide range of usages, from high-speed trains to advanced robotics. To solve these problems and improve the maglev system’s trajectory-tracking performance and robustness, this research proposes a control technique that involves synthesizing a T-S fuzzy controller using the parallel distributed compensation (PDC) method. The controller design is further augmented with a velocity-compensation technique to enable smooth and frictionless ball levitation in a maglev system. The gravitational bias acting on the maglev system is controlled by integrating the feed-forward controller ( $$F_f$$ ) with the PDC-TS fuzzy scheme. The Lyapunov function candidate and linear matrix inequalities (LMIs) are explored to determine the proposed TS fuzzy scheme’s global asymptotic stability. Finally, the effectiveness of the control technique is experimentally evaluated for several test cases using hardware-in-loop (HIL) testing on the maglev system. The results corroborate that the T-S fuzzy control strategy offers robustness and trajectory tracking of the system with stable levitation over the traditional PIV scheme.

Electrode shift can cause significant variability and non-linearity in surface electromyographic(sEMG) signals, which directly impacts the robustness and reliability of EMG based prosthetic control. Hence, to mitigate the misclassification problems in hand gesture due to electrode shift, this study presents a deep multi-view fusion twin support vector machine (DMVF-TSVM) framework which leverages the potential of entropy features to characterize various functional gestures. We harness a variational mode decomposition (VMD) technique and Hilbert-Huang transform (HHT) to decompose non-linear sEMG signals and extract prominent entropy features. Augmenting the benefits of multi-view learning with the Deep Neural Network (DNN) model to extract the shift-invariant deep features, this study adopts a TSVM model to realize a robust hand gesture classification frame work. The primary novelty of this study lies in the deep multi-view learning which fuses the deep features from multiple views and facilitates the models to capture diverse information that are resilient to variations in sEMG signals due to electrode shift. Experimental analysis on publicly available sEMG dataset corroborates that the proposed DMVF-TSVM framework can achieve a superior classification accuracy of 98.34% compared to Deep SVM and Twin-SVM-based models. Furthermore, the proposed framework achieved a mean inter-subject accuracy of 92.60%±1.55% with a Coefficient of Variation (COV) of 1.67% substantiating consistent performance across subjects. Statistical analysis using Friedman test, resulting in p-value of 0.000006, confirmed a significant performance improvement over the competing classifiers.

This paper puts forward a novel deep reinforcement learning control framework to realise continuous action control for a partially observable system. One of the central problems in continuous action control is finding an optimal policy, which can make the agent achieve the control goals without violating the constraints. Although the reinforcement learning technique (RL) is primarily applied for addressing the optimisation problem in continuous action space, the critical limitation of the existing methods is that they utilise only a one-step state transition approach and fail to capitalise on the information available in the sequence of its previous states. Consequently, learning an optimal policy for continuous action space through current techniques may not be effective. Hence, this study attempts to solve the optimisation problem by integrating a convolutional neural network in a deep reinforcement learning (DRL) framework and realise an optimal policy through an inverse n-step temporal difference learning method. Moreover, we formulate a novel convolutional deep deterministic policy gradient (CDDPG) algorithm and present the convergence analysis through the Bellman contraction operator. One of the key benefits of the proposed approach is that it improves the performance of the RL agent by not only utilising information from a one-step transition but also extracting the hidden information from previous state sequences. The efficacy of the proposed scheme is experimentally validated on a rotary flexible link (RFL) system for tracking control and vibration suppression problems. The experimental validation of the proposed scheme on an RFL system highlights that the CDDPG can offer better tracking and vibration suppression features compared to those of the conventional DDPG and the state-of-the-art proximal policy optimisation (PPO) techniques.

EEG signal can capture spatial and temporal shifts in electrical activity of the brain, thereby acting as a prominent biomarker to diagnose seizure. Nevertheless, as EEG recordings are generally high-dimensional and noisy, manual examination of any such abnormalities in EEG patterns to diagnose seizure is a time consuming and tedious task for a neurologist. Therefore, it is important to devise an effective decision support system which can uncover the abnormalities in the EEG signals. In this study, we present an EEG based seizure classification framework which leverages the potentials of variational mode decomposition (VMD) technique and a Bayesian optimized support vector machine (BOSVM) to classify the type of generalized motor and non-motor seizures. Unlike existing methods that rely on statistical or time-frequency features, this study exploits entropy metrics extracted from VMD-decomposed EEG modes for seizure classification. To select the most discriminative features, we harness both neighborhood component analysis (NCA) technique and ReliefF algorithm and apply feature fusion method to train the classifier models. For addressing the class imbalance problem in the dataset, an adaptive synthetic (ADASYN) sampling, which can augment the synthetic samples to minority classes, is adopted. Experiments conducted on a publicly available temple university hospital (TUH) dataset substantiate that the proposed framework achieves an average classification accuracy of 98.37% and offers improved generalization and robustness compared to the state-of-the-art machine learning classifier models.

This paper puts forward a novel multi-agent deep reinforcement learning (MADRL) control framework based on deep deterministic policy gradient (DDPG) algorithm for tracking control of 2 degree-of-freedom (DoF) helicopter system. To handle the nonlinear dynamics and uncertainties in the helicopter system, we formulate a model free data-driven approach which exploits the potentials of reward shaping technique to realise an adaptive and cooperative control policies for efficient trajectory tracking. Specifically, in this study to ensure safe and efficient RL learning without violating the hard constraints of the 2 DoF helicopter, we harness physics informed reward shaping (PIRS) technique which augments domain specific knowledge with data-driven learning. For estimating the pitch and yaw velocities of the helicopter, this study adopts a super twisting observer (STO) based on the second-order sliding mode theory. The key distinguishing features of STO is that it can ensure finite-time convergence with minimal chattering. The efficacy of the proposed MADRL augmented with STO is experimentally validated on a laboratory scale 2 DoF helicopter for several realistic test scenarios including bounded disturbances and uncertain dynamics. The experimental results highlight that the proposed framework can offer better tracking and robustness features compared to conventional state feedback control techniques.

This paper presents a state feedback H ∞ control design for active suspension systems with time-delays using an asymmetric Lyapunov–Krasovskii functional (LKF). Less conservative stabilization conditions are derived in this paper for the design of a state feedback control synthesis through asymmetric LKF in the linear matrix inequalities (LMIs). This investigation aims to exploit the capability of asymmetric LKF to minimize conservativeness and utilizes Jensen’s and Wirtinger’s bounded integral inequalities to handle cross-terms in the asymmetric LKF derivative. One of the key benefits of the asymmetric LKF, compared to symmetric LKF, is its capability to provide less conservative stability conditions by relaxing the condition that all the matrix variables should be symmetric or positive definite. Capitalizing on the potential of asymmetric LKF to realize a less conservative state feedback H ∞ control design, this study evaluates the performance on a multi-objective quarter-vehicle active suspension system (ASS) for various realistic road profiles. Simulation results of bump road surface and random road profiles along with the frequency responses indicate that despite the time delays in the feedback loop the proposed H ∞ controller improves the closed-loop suspension system performance significantly. Quantitative analysis of the chassis acceleration and suspension travel corroborates that compared to the performance of H ∞ controller and predictor-based feedback controller designed using symmetrical LKF, the proposed control scheme reduces the body acceleration and tyre deflection RMS values by 16.5% and 8% when compared to H ∞ control and by 21.5% and 13.5% when compared to predictor-based state feedback control, respectively.

To decode the motion intention from the surface electromyography signals (sEMG) for designing multi-functional prosthetic devices, machine learning (ML) and deep learning (DL) techniques have been widely adopted. Nevertheless, the complexity of the prosthetic hand and the non-stationary characteristics of sEMG signals introduce several practical challenges in adopting ML/DL techniques in the myoelectric prosthetic design. To this end, considerable research attention has been paid to enhance the model reliability, adaptation and robustness. In this article, we present a comprehensive review on the latest advancements in sensing modalities in prosthetics, publicly available datasets, prominent features for classifier training and the feature selection techniques used in the classifier model. Specifically, this article presents a survey on ML and DL techniques for myoelectric hand prosthetic control spanning nearly two decades (2007–2024) by following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and using Web of Science and PubMed databases. This scoping review also elucidates the potentials of different sensing modalities in myoelectric control other than sEMG signal for enhancing the robustness and reliability of the prosthetic device. Moreover, open research challenges and emerging research directions in terms of hardware design, multi-modal sensing, and motion decoding techniques are also critically analyzed to provide insights on future developments.

This paper puts forward a policy feedback based deep reinforcement learning (DRL) control scheme for a partially observable system by leveraging the potentials of proximal policy optimization (PPO) algorithm and convolutional neural network (CNN). Although several DRL algorithms have been investigated for a fully observable system, there has been limited studies on devising a DRL control for a partially observable system with uncertain dynamics. Moreover, the major limitation of the existing policy gradient based DRL techniques is that they are computationally expensive and suffer from scalability issues for complex higher order systems. Hence, in this study, we adopt the PPO technique which utilizes first-order optimization to minimize the computational complexity and devise a DRL scheme for a partially observable flexible link robot manipulator system. Specifically, to improve the stability and convergence in PPO algorithm, this study adopts a collaborative policy approach in the update of value function and presents a collaborative proximal policy optimization (CPPO) algorithm that can address the tracking control and vibration suppression problems in partially observable robotic manipulator system. Identifying the optimal hyper-parameters of DRL using the grid search method, we exploit the capability of CNN in actor-critic architecture to extract the spatial dependencies in the state sequences of the dynamical system and boost the DRL performance. To improve the convergence of the proposed DRL algorithm, this study adopts the Lyapunov based reward shaping technique. The experimental validation on robotic manipulator system through hardware in loop (HIL) testing substantiates that the proposed framework offers faster convergence and better vibration suppression feature compared to the state-of-the-art policy gradient technique and actor-critic technique.

This paper presents a deep deterministic policy gradient-based reinforcement learning (DDPG-RL) controller to address the reference tracking and vibration suppression problem of rotary flexible link (RFL) manipulator. Formulating the control design based on the actor-critic RL method that offers better stability of convergence compared to value based RL method, we synthesize a data-driven control framework that can guarantee precise trajectory tracking while minimizing the vibration of the flexible link. As the RFL system is unstable in open loop, for identifying the dynamical model of the RFL, we build an empirical auto-regressive (ARX) model using the closed loop identification technique. Subsequently, this work also performs the validation test to assess the goodness of the model. Moreover, the efficacy of the DDPG-RL control framework is experimentally validated on a laboratory scale RFL system using the hardware in loop (HIL) testing for precise tracking and robustness. The experimental results substantiate that the DDPG-RL scheme offers precise trajectory tracking and robustness against the external disturbances.

Deciphering and classifying surface electromyography (sEMG) signals is highly essential in rehabilitation robotics, myoelectric prosthetic control, sign languages and human–computer-interface. Researchers have generally focused on utilizing multi-channel sEMG signals for gesture classification. However, in case of amputees, the residual muscles are limited and the gesture classification needs to be reframed with minimum sEMG channels. Moreover, another key problem in the gesture classification system is the nonlinear and non-stationary nature of sEMG signals, leading to poor generalization ability. Hence, to address the aforementioned problems, this study puts forward a novel deep learning classifier framework which leverages the potentials of variational mode decomposition (VMD) technique and a hybrid convolutional neural network–long short term memory (CNN–LSTM) classifier model to recognize the hand gestures from single channel sEMG signals. Collecting the sEMG signals from forearm muscles of 25 intact subjects for ten functional and grasping actions, this work implements a VMD technique to identify the prominent frequency modes in the sEMG signals. From the decomposed modes of sEMG signals, the prominent intrinsic mode functions (IMFs) are extracted through spectral analysis to minimize the computation burden on the hybrid classifier model. Furthermore, as the muscle contractions result in substantial temporal dependencies, this work exploits the potentials of CNN and LSTM networks and extracts the spatiotemporal features of the sEMG signals for various hand gestures. The experimental results corroborate that the proposed classifier framework can achieve an average classification accuracy of 98.04% and provide 3% improvement in the classification accuracy compared to conventional CNN classifier.

To improve the classification accuracy of hand movements from sEMG signals, this paper puts forward a unified hand gesture classification framework which exploits the potentials of variational mode decomposition (VMD) and multi-class support vector machine (SVM). Acquiring the sEMG signals from 25 intact subjects for ten functional activities in real-time, we implement a non-recursive adaptive decomposition technique to sEMG signals and perform power spectral analysis to identify the dominant narrow-band intrinsic mode functions (IMFs) that contain prominent biomarkers. Subsequently, to compute the optimal feature vectors from a set of entropy measures, this work investigates the performance of two techniques namely minimum redundancy and maximum relevance (MRMR) technique and kernel principal component analysis (kPCA). After extracting the optimal set of entropy features, the proposed approach implements a multi-class SVM based on one-vs-one (OVO) strategy to classify the hand gestures. The performance of the multi-class SVM compared with those of the K-nearest neighbor (KNN) and naïve bayes (NB) classifiers highlight that multi-class SVM offers superior performance with an average classification accuracy of 99.98%. Moreover, for statistical analysis of the experimental results, this work performs Friedman test to analyze the significance of the SVM, KNN and NB classifier performances. Finally, the performance comparison of the proposed approach with those of the state-of-the-art techniques highlights the superiority of the proposed framework to improve the hand gesture classification accuracy.

This paper puts forward a novel Hybrid Coyote Optimization-based Big Bang Big Crunch (HCOB3C) algorithm to design an optimal multi-objective fuzzy control framework applied to Active Suspension Systems (ASS). The suspension system in vehicles is an inherent component that is responsible for yielding passenger comfort and ensuring vehicle stability. Since ASS is a multi-objective, constrained non-linear system, the linear controllers will yield suboptimal results because of the so-called bode sensitivity integral problem. Hence, to handle the non-linearity and constraints in the ASS, we present a constrained multi-objective fuzzy controller optimized using the HCOB3C algorithm. The motivation for the proposed hybrid optimization algorithm is that the conventional Big-Bang Big Crunch Optimization (B3CO) and Coyote Optimization (CO) suffer from two major limitations namely 1. Imbalance between exploration and exploitation and 2. Slow convergence respectively. Hence, we utilize the CO to tune the parameters of B3CO to realize optimal actuator force that can offer precise suspension travel and minimize the chassis vibration even in the case of uneven road profile. The performance of the proposed scheme is experimentally validated on a quarter car ASS system for several realistic road profiles. The experimental results substantiate that the proposed scheme can minimize the vehicle vibration by around 41.6

To address the nonlinear stabilization problem and improve the tracking control feature of ball on plate system (BPS), this paper puts forward a novel Takagi Sugeno (TS) fuzzy control augmented with the current cycle feedback iterative learning control (CCF-ILC) scheme. According to Bode's sensitivity integral, the performance of linear controllers is always a trade-off between reference tracking and robustness. Hence, to deal with the so-called 'waterbed' effect, this work exploits the capability of TS fuzzy to handle the nonlinear dynamics and synthesizes a learning control scheme based on current iteration error to capitalize the information rich error signal for enhancing the robustness and trajectory tracking features. The global asymptotic stability of the proposed TS fuzzy augmented ILC scheme is proved using the Lyapunov function and linear matrix inequalities (LMIs). Moreover, the monotonic convergence of ILC is presented based on the singular value condition. For identifying the rolling mass from the video stream, a background subtraction algorithm based on thresholding technique is implemented. Finally, the robustness and tracking features of the proposed scheme are evaluated on a two degrees of freedom (DoF) laboratory scale BPS system through hardware in loop (HIL) testing for three realistic test cases. The tracking performance quantified using the root mean square error (RMSE) and power spectral density plot corroborates that the proposed scheme can offer better setpoint tracking and robustness feature compared to state-of-the-art fuzzy and ILC control techniques implemented on BPS.

Parkinson's disease (PD) is a progressive, debilitating neurological movement disorder that affects the person's muscle control, movement, speech, cognition and dexterity. For diagnosing PD in a clinical setting, in addition to the neurological examinations, clinicians use the unified Parkinson disease rating scale (UPDRS) to assess the motor and non-motor impairments. Such a clinical assessment highly depends on the experience and expertise of the clinicians, and it may result in biased evaluation. Hence, to assist the clinicians, we put forward a gait analysis-based deep convolutional neural network (DCNN) framework which leverages the potentials of variational mode decomposition (VMD) technique with the recurrence plots (RP) to enhance the PD severity classification performance. Specifically, transforming the VMD modes of vertical ground reaction force (VGRF) time series data into two-dimensional texture images to capture the temporal dependency, this work trains the DCNN classifier through recurrence images for its ability to extract the discriminative features among the PD severity levels. For evaluation, this study utilises the VGRF dataset of 93 PD subjects and 73 healthy controls from Physiobank for three different walking tests. Consequently, utilising VMD, RP and DCNN in a unified framework, this investigation shows that the PD severity rating can be significantly enhanced through DCNN model that is trained using RP of dominant intrinsic mode functions (IMFs). The novelty of the proposed framework lies in identifying the prominent gait biomarkers through dominant IMFs from power spectral analysis for reducing the computational burden of DCNN. Moreover, to handle the data over-fitting issue in the classifier, L2 regularisation technique, which penalises the weight parameters of the nodes, is used in combination with the dropout layer. Experimental results underscore that the proposed VMD-RP-DCNN architecture can address the spectral overlapping issue in VGRF decomposition and achieve an average PD severity prediction accuracy of 98.45%.

This paper puts forward a novel deep reinforcement learning control using deep deterministic policy gradient (DRLC-DDPG) framework to address the reference tracking and vibration suppression problem of rotary flexible link (RFL) manipulator. Specifically, this study attempts to address the continuous action space DRLC problem through DDPG algorithm and presents a Lyapunov function based reward shaping approach for guaranteed deep reinforcement learning (DRL) convergence and enhanced speed of training. The proposed approach synthesizes the hard and soft constraints of the flexible manipulator as a constrained Markov decision problem (MDP) and evaluates the performance of DRLC-DDPG framework through hardware in loop (HIL) testing to realize precise servo tracking and suppressed vibration of the flexible manipulator. For identifying the dynamical model of the RFL, an empirical Auto-Regressive eXogenous (ARX) model using the closed loop identification technique is built. Moreover, to extract the true states (servo angle and deflection angle) from the actual measurements, which typically have the influence of sensor noise, an adaptive Kalman filter (AKF) is augmented with the DRLC scheme. The experimental results of DRLC-DDPG scheme compared with those of the model predictive control (MPC) for several test cases reveal that the proposed scheme is superior to MPC both in terms of trajectory tracking and robustness against the external disturbances and model uncertainty.

This paper puts forward a novel entropy features based multi-class SVM classifier framework to predict the limb movement of the transradial amputees from the surface electromyography (sEMG) signals. The major challenges with the sEMG signal are nonlinear and non-stationary characteristics and susceptibility to noise. Consequently, a robust and an effective feature extraction framework which is invariant to force level variations is central in sEMG based prosthesis control. To address the aforementioned challenges, this study leverages the potential of variational mode decomposition (VMD) technique to identify the prominent frequency modes of the sEMG signals, and performs the spectral evaluation of the decomposed sEMG modes to identify the dominant ones to extract the entropy features. Subsequently, we evaluate the efficacy of four nonlinear optimal feature selection techniques and identify the prominent entropy features to train the multi-class SVM model that can predict the gestures. Specifically, to handle the nonlinearly separable input data, this study implements a kernelization named a radial basis function (RBF), which has good generalization and noise tolerance features. The efficacy of the proposed framework is tested using the publicly available datasets that contain gestures from transradial and congenital amputees for functional gestures. Experimental results obtained for various gestures with dynamic force levels underscore that the proposed framework is highly robust against the force level variations and can achieve a classification accuracy of 99.07%.

This paper presents a novel adaptive augmented linear quadratic integral (LQI) controller for a 2 degrees of freedom (DoF) helicopter to improve the closed-loop performance of the baseline controller even when the system contains unmodeled dynamics and exogenous disturbances. Although the LQI controller is widely used in aircraft control applications for its excellent reference-tracking capability and inherent robustness, the performance of the baseline LQI controller degrades when the aircraft model has unmodeled dynamics and external disturbances, which are inevitable in aircraft applications. Hence, to deal with the parameter uncertainty and unmeasured disturbances, we present an adaptive control framework using a model reference adaptive control (MRAC) scheme combined with the baseline LQI controller. The key aspect of the proposed control scheme is that the robustness of the closed-loop control system to deal with the matched uncertainty is enhanced by tracking error augmentation with the control law. Moreover, this work proves the asymptotic stability of the closed-loop system using the inverse Lyapunov function and Barbalat's lemma. The efficacy of the adaptive augmented LQI is experimentally validated through hardware in loop (HIL) implementation on a laboratory scale 2 DoF helicopter workstation for several real-time test cases. Experimental results accentuate that the adaptive augmented control scheme can significantly improve not only the reference-tracking capability but also the robustness of the closed-loop system.

To deal with multiple constraints of vehicle active suspension system (ASS) including road handling and passenger safety, this paper presents an optimal linear quadratic regulator (LQR) approach which employs bat algorithm (BA) for selection of optimal state and input penalty matrices of LQR. We formulate the conflicting control objectives of ASS, namely, ride comfort and passenger safety as a multi-constraint optimization problem and employ the BA for weight selection of LQR. The key advantage of the proposed approach is that the local optima problem is avoided by utilizing the frequency tuning and random walk technique in BA. The performance of the proposed approach is experimentally tested using hardware in loop (HIL) testing on a quarter car ASS for realistic road profiles. Moreover, the performance is benchmarked against grey wolf optimization tuned LQR. Experimental results assessed based on ISO 2631 standards highlight the significant improvement in the ride comfort and passenger safety.

Among the many vital parameters of human body, breath rate monitoring is paramount to detect the symptoms of several respiratory diseases such as sleep apnea syndrome, chronic obstructive pulmonary disease (COPD), and asthma. Hence, this paper puts forward a novel framework to measure the respiration rate (RR) using the reflective type photoplethysmogram (PPG) signals acquired using an inexpensive and easy-to-use wearable de -vice. Extracting the respiration induced amplitude variations (RIAV) from PPG signal using the incremental merge segmentation (IMS) algorithm, the proposed approach shows that robust RR estimation in realtime is feasible through low cost Cortex-M4 microcontroller. Using the sliding window approach augmented with adaptive thresholding technique that can deal with motion artifacts, we show that robust RR estimation is viable using reflectance type PPG, which is largely unexplored. Moreover, to handle the non-uniform nature of RIAV signal, this work employs uniform interval interpolation technique and removes the non-respiratory frequencies using finite impulse response (FIR) band-pass filter. Through the fast Fourier transform (FFT) analysis of regular interval RIAV sequence, the dominant frequency corresponding to the RR is extracted. The performance of the proposed scheme is validated not only on two publicly available PPG datasets but also on a custom designed experimental setup integrated with the smartphone. The experimental results highlight that our approach can achieve a RR estimation accuracy of 1 bpm deviation against the ground truth.

To improve the trajectory tracking and robustness of closed-loop servo system against the model perturbation, this paper presents a novel norm optimal iterative learning control (NOILC) scheme combined with proportional velocity (PV) feedback control. It is well known that the feedback controller performance is always limited due to the so-called Bode sensitivity integral, which states that the feedback controller performance is always a trade-off between the reference tracking and the disturbance rejection. Hence, to address this trade-off called “waterbed effect”, we synthesize a NOILC scheme, which can significantly improve the tracking performance by learning the system dynamics through the past tracking errors and the control effort. Formulating the ILC design as an optimization problem, we determine the optimal learning filters and present the hardware in loop testing (HIL) validation of the proposed scheme on a servo motor. Experimental results substantiate that the NOILC combined with PV can significantly reduce the tracking error and enhance the transient and steady-performance.