Yaw stability control systems are one of the approaches implement for lateral dynamics motion control of road vehicles. An important variable that we need to control in order to achieve vehicle lateral stability is the vehicle yaw rate by proposing an upper controller that is based on the Sliding Mode Control (SMC) technique. The objective for the controller is to control the vehicle yaw rate to a reference trajectory coming from the nonlinear single-track vehicle model assuming that the desired side slip angle is equal to zero. The proposed SMC-based upper controller will adjust the steering angle of the front wheel through the steer-by-wire system for yaw motion stabilization control to account for the system's nonlinearities and external disturbances. Simulation results show robustness of stability and performance using the SMC logic with the ability to react to the dynamic behavior experienced in different driving situations including a lane change and emergency maneuver.
Background: Coordinating differential-drive mobile robots for landmark coverage is challenging due to non-holonomic dynamics, clutter, and sparse rewards. Standard multi-agent RL pipelines often show unstable learning and inconsistent completion in this setting. Methodology: We adopt a centralized-training, decentralized-execution actor-critic without inter-agent commu nication. Our replay-centric design combines a tagged buffer that up-samples goal-reaching transitions and an offline replay initialization that seeds early learning with curated trajectories. Dynamic task assignment uses the Hungarian algorithm during training and evaluation, and we benchmark against uniform replay and established variants. Results: In a cluttered six-robot arena, the approach improves training stability relative to uniform replay. Convergence is faster and requires fewer updates to reach consistent success. Coverage efficiency increases as landmarks are reached earlier across runs. Collisions per episode decrease without adding communication or architectural changes. Multi seed evaluations show gains that persist with narrow confidence intervals. Train-evaluation gaps shrink on unseen maps, indicating improved generalization. Ablations attribute complementary ben efits to the tagged and offline components. Performance remains competitive with prioritized and hindsight replay baselines under matched budgets. Computational overhead is small because sam pling logic changes while network sizes do not. Conclusions: Focus ing on replay design substantially stabilizes multi-agent learning for differential-drive coordination. The pipeline integrates cleanly with standard CTDE implementations and supports practical deployment in coverage tasks.
This paper presents a regularized Extreme Learning Machine (ELM) framework for identifying nonlinear dynamic systems affected by multicollinearity, with application to a Hammerstein-structured model of a Continuous Stirred Tank Reactor (CSTR). The model architecture employs a single-hidden-layer feedforward network (SLFN) for the static nonlinear block, and an autoregressive linear dynamic block whose order is determined using a Lipschitz quotient criterion. Traditional ELM models are known to suffer from instability when lagged input variables are highly correlated, a common occurrence in block-oriented system identification. To address this, the study investigates enhanced variants of ELM incorporating regularization, namely Ridge-ELM and Liu-ELM, which introduce biasing parameters to improve numerical stability and generalization. The proposed regularized ELM variants are evaluated against traditional ELM using performance metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). Results show that Ridge-ELM and Liu-ELM significantly reduce parameter variance and improve predictive performance on datasets. Additionally, confidence intervals and condition number analysis demonstrate improved robustness in the presence of multicollinearity. Cross-validation is used to tune hyperparameters, and the Diebold-Mariano test confirms that the improvements are statistically significant. This approach offers a computationally efficient, scalable solution for robust nonlinear system identification in multivariate chemical processes and beyond.
Accurate prediction of maize seed vigor is essential for achieving uniform germination, stable crop yields, and resilience in farming systems; however, traditional laboratory methods, such as germination assays and staining, are destructive, slow, and unsuitable for large-scale deployment. Hyperspectral imaging has shown promise as a noninvasive tool, but existing approaches often rely on spectral data alone and overlook the influence of environmental factors, limiting their usefulness in real agricultural settings. This study introduced HiveFusion, a biologically inspired framework designed to integrate hyperspectral reflectance data with environmental parameters, specifically temperature, humidity, and soil moisture, to improve the reliability of vigor prediction. The proposed system follows a structured process: denoising and correcting spectral signals, normalizing them using an environmentally aware scaling method, selecting informative spectral bands, compressing redundant features, and generating compact latent representations. These spectral features are then combined with environmental input through a mid-level attention-fusion mechanism that assigns importance to each modality, enabling the model to adapt dynamically to different conditions. A dual network, consisting of a lightweight convolutional branch for hyperspectral data and a multilayer perceptron for environmental data, performs final classification into three categories: high, moderate, and low. Experiments on a dataset of 17,724 seed images from three Ghanaian maize varieties demonstrated that HiveFusion achieved 91% accuracy, 0.90 F1-score, and 0.93 ROC-AUC, outperforming both unimodal and simple fusion baselines. Although environmental features were synthetically generated, the results highlight the value of incorporating contextual factors. Future work will validate the model with real sensor data to ensure broader applicability.
Variable Frequency Drives (VFDs) serve as essential elements for modern industrial operations which focus on enhancing energy efficiency and decreasing operational expenses. These systems encounter major stability issues when they operate under adverse conditions which include sudden load changes and power disturbances and delays in signal processing and mechanical system responses. Sliding mode control (SMC) has proven to be an effective solution because it provides adaptable monitoring techniques which also maintain system stability. The research study delivers its main contribution through the implementation of linear matrix inequality (LMI) method within the SMC framework to enhance stability in multi-agent VFD systems. The proposed technique operates to direct the switching functions of the DC link DC-DC CUK converter in the subject system. The subject system exists in mathematical form which allows researchers to study its behavior when exposed to standard input testing signals and its stability characteristics. The system maintains its stability during quick load variations which proves that the control method produces better results for systems that manage speed and torque in motor groups. The system stability during fast load variations proves that the control method produces better results for motor group speed and torque control systems which makes the technology suitable for transportation systems and renewable energy systems and other applications. Standard VFDs present challenges because they require specific motor types and operate with various communication protocols and expensive components. The authors indicate that it should be investigated to gauge the feasibility of having more advanced control algorithms incorporated with the suggested control system; enhancing its adaptability and performance control.
Semi-active suspension systems have emerged as an attractive alternative to fully active suspensions because they offer a superior capacity to improve vehicle ride comfort and handling performance with significantly lower energy consumption. Conventional semi-active control strategies, however, such as skyhook damping, often cannot accommodate the nonlinear and time-varying dynamics of vehicles in operation under impulse or severe road disturbances. In this context, an intelligent smart-damper controller is proposed in this paper by incorporating a Modified Fuzzy Adaptive Fuzzy Logic Control framework in a half-car suspension model. In the developed controller, the effective damping force is adaptively tuned using real-time measurements of body acceleration and velocity to achieve enhanced dynamic robustness. The research contribution is the development of an adaptive, computationally efficient semi-active control law that is capable of achieving superior performance over conventional skyhook damping in the case of highly transient excitations. For this purpose, a comprehensive simulation study has been carried out to evaluate the passive, skyhook, and MFAFLC suspensions for identical Gaussian impulse road profiles. The MFAFLC system results in substantial improvements over passive suspension by reducing peak body displacement by 48.6%, pitch angle by 42.1%, and vertical acceleration by 55.7%, while reducing settling times by 35–50%. MFAFLC thus offers a further improvement of 12–25% over skyhook control for most performance indices. These results illustrate that MFAFLC-based smart damping promises to be a more adaptive and effective solution for semi-active vibration control in vehicles subjected to unpredictable road disturbances.
This paper introduces an adaptive neural network (ANNs) based Radial Basis Function (RBF) for robotic manipulators in order to solve unknown dynamics, external disturbances and dead-zone compensation, thereby enhancing the accuracy of tracking control. To address these problems, two Radial Basis Function neural networks (RBFNN) are designed: the RBFNN is employed to estimate the nonlinearities in the actuator, and the other RBFNN is utilized to compensate for dead-zone in the system’s feedforward channel. The RBFNN is a method that delivers good control performance for systems with uncertain models because of its fast-learning algorithm and strong approximation capability. The design of online adaptive control training laws and dead-zone estimation is caried out using Lyapunov stability theory together with approximation theory. Besides, a robust control plays an auxiliary controller to guarantee the robustness and stability under various environments. Simulation results demonstrate the proposed controller achieves significant improvements in tracking accuracy, reducing the steady-state tracking error by up to 90% comparing with baseline methods, while ensuring smooth convergence and robust performance under parameter variations and disturbances. The dead-zone compensation effectively eliminates oscillations and overshoots that appear in the uncompensated case. These results validate the effectiveness and reliability of the proposed approach for high-precision robotic tracking control.
Radio frequency RF impairments, which consist of thermal noise, phase noise, and non-linearity losses in RF transceivers, heavily degrade the performance of multiple input, multiple output (MIMO) communication systems. RF impairments, including thermal noise, phase noise, and non-linearity in transceivers, significantly degrade the performance of multiple input multiple output (MIMO) communication systems. This study investigates how these impairments affect key performance metrics such as signal-to-noise ratio (SNR), bit error rate (BER), and system capacity. Through extensive Python-based simulations, I analyze the degradation caused by RF impairments and evaluate mitigation strategies to enhance system reliability. The findings provide insights into optimizing MIMO system design for improved spectral efficiency and robustness in adverse conditions. This work has very strong improvements with regard to the performance of MIMO communications by mitigating RF impairments, which lower its performance. These key contributions come in the form of 95 Mbps throughput, latency as low as 10 ms, and high spectral efficiency of 7 bps/Hz. Other advanced techniques also achieve high energy efficiency to 4.5 bits/Joule and reduce Error Vector Magnitude to -35 dB, thus showing enhanced signal quality and reliability. Moreover, there is a reduction in the packet loss rate to 0.005% and interference mitigation by 30 dB to ensure performance in adversarial conditions. The impact of thermal noise, phase noise, and non-linearity is quantified in this research, which provides valuable in-sight into the improvement of MIMO system design and mitigation of the RF impair-ments' adverse effects.
Recent advances in Unmanned Aerial Vehicles (UAVs) for Precision Agriculture have evolved from performing passive tasks, such as monitoring, mapping, and inspection, to active tasks including harvesting, crop sampling, and insect trap deployment. These new applications were carried out through the integration of robotic arms attached to the base of the platform, thereby forming unmanned aerial manipulators (UAMs). However, UAMs encounter several challenges related to the inverse kinematics problem ensuring a precise trajectory and motion under platform constraints and dynamic coupling. In this paper, the objective is to achieve a precise grasping of a plant sample for ex-situ analysis using a quadcopter equipped with a two-degree-of-freedom robotic arm. The main contribution is the development of an accelerationlevel quadratic programming (QP) approach to solve inverse kinematics and to minimize the end-effector tracking error by incorporating joint constraints, self-collision avoidance, and UAV orientation constraints, thereby exploiting the system’s redundancy. The coupled dynamics of the vehicle are formulated using the Euler-Lagrange formalism and a Sliding Mode Controller based on the super twisting algorithm is employed to ensure trajectory robustness and accuracy, while reducing chattering effects. The simulation results demonstrate high tracking accuracy the precision of (RMSEqp = 10−6m) in end-effector trajectory accuracy, with convergence achieved in 3 seconds and a 88% reduction in chattering compared to conventional SMC methods. Comparative analysis against hierarchical QP methods confirms superior trajectory precision and disturbance rejection. These contributions will enable crop collection tasks to be carried out accurately and with high stability and robustness.
This paper proposes a Novel Adaptive Fuzzy Neuro Sliding Mode Controller (NAFNSMC) for a Coupled Tank System (CTS), aiming to achieve adaptive control of unknown dynamic nonlinear systems under experimental conditions. The CTS exhibits strong nonlinearities and uncertainties arising from sensor noise, parameter variations, variations in output valve characteristics, and significant time delays. The proposed control architecture integrates two synergistic components. The first component is an adaptive controller that utilizes a Radial Basis Function Neural Network (RBFNN) to approximate the adaptive control law, featuring an adaptive updating mechanism to compensate for RBFNN approximation errors. The second component is a sliding mode controller whose parameters are updated in real-time via a fuzzy inference mechanism to enhance robustness and improve convergence speed. Both control laws are derived within the framework of Lyapunov stability theory, ensuring closed-loop stability under all operating conditions. The stability is guaranteed under bounded neural approximation errors. The proposed NAFNSMC is experimentally validated and compared with an adaptive neural sliding mode (ANSMC) and a PID controller tuned using the standard Ziegler–Nichols II (PID Z–N II) method. Experimental results demonstrate that the NAFNSMC achieves superior tracking performance, particularly under significant external disturbances, with a 54.3% improvement in tracking accuracy compared to the ANSMC. However, a limitation of this study is that the controller parameters are selected empirically through trial and error, which may hinder the achievement of optimal parameter values.
This paper addresses the trajectory tracking problem of an omnidirectional mobile robot (OMR) operating undermodeling uncertainties and external disturbances. The proposed controller integrates a Proportional–Integral (PI) controller with Dynamic Surface Control (DSC). An Adaptive Neuro-Fuzzy In-ference System (ANFIS) is employed to simplify implementation by reducing the dimensionality of the controller input data. Based on the dynamic model, a DSC strategy is designed to enhance robustness and tracking accuracy. Subsequently, using offline training data, a structural transformation technique is derived to embed the DSC into a PI controller for each input channel. The proposed approach inherits the simplicity of the PI controller and the robustness of DSC, achieving a decentralized SISO controller as an alternative to the conventional centralized DSC structure. Simulation studies are conducted under two scenarios:without disturbances and with disturbances. The results show that the proposed controller improves position tracking accuracy by approximately 40% compared to a conventional PI controller. In addition, a comparison with a robust adaptive controller is presented to demonstrate the efficiency of the proposed approach. These findings highlight the potential of the ANFIS-based PI–DSC controller as a practical solution for simplifying controller design and enhancing the robustness of OMR systems under adverse environmental conditions.
Speed sensorless control in three-phase induction motors relies on accurate observer‑based speed estimation, making it susceptible to current sensor faults. Thus, incorporating an FTC mechanism is crucial to maintain robustness and reliability under fault conditions. This study contributes to provides a significant advancement in the development of resilient control systems for electric drives, addressing critical challenges in maintaining performance under fault conditions. This study develop a FTC strategy integrated with sensorless speed control for an induction motor hardware platform. A disturbance observer (DO) approach is employed to simultaneously estimate both rotational speed and disturbance signals, enabling the distinction between current sensor faults and load torque variations. Experimental-based evaluations were performed using integration between MATLAB/Simulink and hardware. The performance of the proposed algorithm was evaluated through three tests: the Set‑Point Change Test, the Torque‑Load Variation Test, and the Current‑Sensor Fault Test. Compared with a baseline DO-based sensorless controller without FTC, the proposed method prevents fault-induced degradation (e.g., loss of steady-state tracking/hunting behavior) by reconstructing the corrupted current channel and maintaining stable closed-loop operation during the fault. Under a step bias fault, the proposed scheme achieves a current recovery time of approximately 1.56 s, after which the speed response returns to the steady-state tracking band without current surge. The results demonstrate that the proposed method effectively compensates current-sensor faults in measurable terms thereby improving the robustness and reliability of DO-based sensorless induction-motor drives.
Trajectory planning on an eye-in-hand robotic manipulator with a cluttered heterogeneous environment is crucial to grasping success. This study proposes a hybrid model that combines deep learning-based environment detection with reinforcement learning (RL) to determine the optimal route. You Only Look Once (YOLO) was proposed in this system to incorporate deep reinforcement learning by optimizing the environment assessment. Merging YOLO with deep reinforcement learning optimizes environmental assessment by inserting a reward subprocess. YOLO assesses the environment for detection and recognition; after identifying targets or obstacles, RL can then reward or punish the agent to achieve the highest score. The entire process is encapsulated within the environmental cycle, which incorporates a state reward that is analyzed by deep learning. On the agent side, it is also driven by reinforcement learning. In this order, each agent and environment reach their optimal values, resulting in a 11-17% decrease in episode convergence time.
This paper investigates the impact of steady-state error minimization on the performance of numerical optimization techniques in linear automatic control systems, introducing a novel framework that integrates advanced genetic algorithms and machine learning to enhance controller tuning It highlights the significance of selecting appropriate test signals to generate quality system responses, which directly affects stability and reliability. Various optimization techniques are discussed, including classical methods and modern algorithms such as genetic algorithms and machine learning. Special attention is given to astatic control, which minimizes static errors and enhances controller reliability. Experimental results reveal that optimizing for one signal type can significantly diminish performance for another type. The paper introduces trade-offs that facilitate simultaneous consideration of performance responses to various stimuli. The conclusions underscore the importance of carefully selecting test signals and provide recommendations for automatic control practitioners, ultimately leading to improved reliability and efficiency in systems under dynamic conditions.
Temperature regulation is crucial for crop yield optimization in controlled environment agriculture, yet achieving such accuracy is challenging due to system nonlinearities and external disturbances. Since H_(∞ )control is an established theory, its experimental validation on low-cost hardware for agricultural systems remains limited. This paper presents robust control for a nonlinear system, targeting an internal temperature of growth chamber agriculture. Moreover, the primary contribution is the demonstration of a systematic and practical methodology for designing, implementing, and validating an H_(∞ ) controller on an Arduino-based growth chamber prototype, bridging the gap between complex control theory and accessible implementation. A simplified linearized thermal model was derived from a lumped parameter approach using energy balance equations. A second-order weighting function was systematically designed using loop-shaping principles to guarantee robust performance against unmodeled dynamics and sensor noise. The resulting controller was synthesized in MATLAB and deployed on an Arduino Mega microcontroller for experimental testing. Simulations predicted high-precision tracking with a Root Mean Square Error (RMSE) of 0.037 °C and an Integral Absolute Error (IAE) of 0.70. Subsequent experimental validation under real-world conditions confirmed the controller's efficacy, achieving stable temperature regulation within ±2 °C of the set point. The experimental validation yielded an RMSE of 1.04 °C and an IAE of 0.924, highlighting a notable but analyzed performance gap between the idealized simulation and the physical implementation. The results of this work were also compared with MPC and PID controllers, showing the proposed approach demonstrated satisfactory performance and confirming the robustness and stability of the control strategy in practical conditions. This work concludes that the H_(∞ ) framework provides a computationally efficient pathway to achieving robust temperature control on accessible hardware, making advanced control techniques more feasible for distributed agricultural applications.
Cancer remains a major global health burden, and early detection is critical for reducing mortality. Conventional machine learning often depends on centralized data, which raises privacy and data-sharing concerns. Federated learning (FL) offers a solution by enabling collaborative model training across institutions without sharing raw data, safeguarding privacy while improving generalizability. The contribution of this review is to systematically analyze FL applications in cancer detection, highlighting their strengths, limitations, and future directions for clinical translation. The review was conducted through a structured search of Scopus, focusing on peer-reviewed, open-access articles in English. A total of 42 studies were included, spanning publications from 2020-2024. Eligible studies were screened based on relevance to oncology and FL, with data extraction covering cancer type, data source, FL methods, models, data types, performance metrics, key findings, and research gaps. Results show that FL has been applied to a wide range of cancer types including breast, lung, prostate, colorectal, brain, cervical, melanoma, and multi-cancer datasets. Data sources involve both multi-center collaborations and public datasets, while methods include horizontal, vertical, hybrid, and variants such as FedAvg, FedProx, and transfer learning. Models range from CNNs, ResNets, and UNet derivatives to transformers and ensembles. Reported metrics indicate high performance comparable to centralized learning. Key findings highlight privacy preservation and robust generalization, but research gaps remain in dataset size, heterogeneity, validation, computational cost, and interpretability. In conclusion, FL shows strong promise for collaborative cancer detection, yet future studies must address scalability, data diversity, and transparency to support real-world clinical adoption.
This paper investigates finite-time stability (FTS) and finite-time synchronization (FTSY) in discrete generalized Gray Scott reaction-diffusion systems (GS-RDs). We develop theoretical frameworks for analyzing these properties and propose explicit control laws to achieve FTSY between master and slave systems. The approach leverages Lyapunov functions, discrete summation by-parts, and bounds on nonlinear terms to establish conver gence within a finite settling time. Numerical simulations validate the theoretical results, demonstrating synchronization within an estimated time of T2 = 29 seconds under specified control strategies. The study highlights practical implications for real-time control systems and communication networks, while acknowledging limitations such as parameter sensitivity and computational complexity.
The tracking control problem for a gun platform plays an important role in military tasks. However, in obstacle-rich space, the system faces the energy minimization by adjusting the control signals in case of tracking a moving target. This paper presents a solution for applying optimal control to adapt to changes in model parameters describing the electromechanical system of the artillery platform in the process of tracking mobile targets in the air. Moreover, the unification of tracking performance and optimal control is necessary to be guaranteed to improve control performance in the gun platform system. To address the challenge of constructing the tracking error model, the Linear Tracking Quadratic (LQT) is employed to develop the optimal control scheme. According to the corresponding approximation, a novel optimal tracking algorithm is presented for a Gun Platform with the tracking objective of following a moving target. Extensive simulation results of the cost function with comparisons demonstrate the robustness and efficiency of the proposed method.
This paper introduces an Enhanced Generalized Modified Repetitive Control (EGMRC) scheme in which the compensator is derived directly in the continuous-time domain, avoiding conversion to a two-dimensional hybrid model. The proposed method is implemented for angular position tracking of a 1-DOF lower-limb robotic exoskeleton driven by a periodic reference. The closed-loop dynamics are expressed in an error-state form to explicitly capture the tracking objective. Controller gains are determined via linear matrix inequality (LMI) conditions to ensure asymptotic stability, while performance is shaped by tuning the relative contributions of the control and learning actions and by adjusting the internal-model filter cut-off frequency. Numerical studies verify the proposed approach and show that the most accurate tracking is obtained when the learning-action gain exceeds the control-action gain. The results further suggest that increasing the internal-model cut-off frequency generally improves tracking accuracy. Compared with generalized repetitive control (GRC), EGMRC achieves superior performance in the evaluated metrics, with the exception of settling time.
Cut-in maneuvers pose a major challenge for Adaptive Cruise Control (ACC) systems because they require rapid re-identification of the lead vehicle and immediate adjustment of longitudinal control to maintain safety. Conventional ACC controllers often struggle to respond quickly and smoothly under these abrupt disturbances. This study proposes the use of the Crow Search Algorithm (CSA) to optimize key ACC control parameters, enabling improved adaptability and stability during cut-in events. CSA is employed to tune the controller gains through an iterative search process that balances exploration and exploitation. Simulation experiments are conducted to evaluate the influence of iteration numbers and population sizes on optimization performance. The optimized ACC system is validated through a cut-in scenario, assessing relative distance error, velocity error, and recovery behavior after the disturbance. The results show that CSA significantly enhances the controller’s ability to reduce transient errors, stabilize inter-vehicle distance, and restore speed in a smooth and safe manner. These findings demonstrate that CSA is a promising metaheuristic for improving ACC robustness in highly dynamic traffic environments. Future extensions include testing in multi-scenario conditions, incorporating cooperative vehicle-to-vehicle communication, and integrating hybrid optimization methods to further enhance real-time performance.
