Explore the vast range of titles available.

Focusing on a non-parallel distributed compensation (non-PDC) control scheme, this paper aims to derive a stabilization criterion for continuous-time Takagi-Sugeno (T-S) fuzzy delayed systems in the presence of the asynchronous fuzzy phenomenon. To achieve this goal, the paper presents a scheme for formulating fuzzy-dependent stabilization conditions that can handle delayed states, aiming to obtain less conservative performance and reduce computational complexity. Specifically, a control synthesis method is derived that is capable of designing both the free-weighting matrix and the congruent transformation matrix, relying on the asynchronous fuzzy basis function. Furthermore, the asynchronous fuzzy phenomenon is addressed by transforming the asynchronous fuzzy basis functions into the errors between the original and asynchronous fuzzy basis functions, and incorporating these errors into the fuzzy-dependent stabilization criterion.

This paper proposes an end-to-end pipeline to detect broken eggs in a holder without extensive training, employing a two-step image segmentation and processing approach using saliency scores, all without relying on a large amount of labeled data. The process begins by inputting an egg image with text prompts into Grounding DINO, which returns an egg bounding box. This is followed by the segment anything model (SAM), which extracts the egg’s segmented region. The segmented region is then divided into two crucial components for detection: a binary mask image and a background-removed egg image. The innovation in our method lies in using the saliency score of the estimated anomaly region by employing image processing techniques to effectively distinguish between intact and broken eggs. To validate our approach, we compare it to well-known models such as SVM, XGBoost, and YOLOv8, and we also conduct zero-shot experiments with CLIPSeg, Florence-2, and SAA. In our experimental setup, we utilize 50 egg holder images, each containing both intact and broken eggs. We carefully cropped and processed 30 eggs (arranged in a 6×5 grid) from each holder, resulting in a comprehensive testing dataset totaling 1,500 images. Our results demonstrate the robustness of our method, achieving an impressive 99.56% accuracy in detecting both intact and broken eggs. This breakthrough promises significant advancements in the field of broken egg detection, with broad applications across diverse industries, including food safety, quality control, and automated packaging systems.

This paper aims to study the local stabilization problem of discrete-time fuzzy systems via a quadratically structural approach. The proposed methods tackle the conventional analysis deriving stabilization conditions in multiple summations by relaxed stabilization conditions in a quadratic structure. Moreover, an attempt to utilize the inherent properties of the quadratic structure is first introduced to construct a feasible problem in the form of linear matrix inequalities. These endeavors allow optimization problems to maximize a region of attraction to be achieved with relaxation. Finally, the effectiveness of the proposed methods is discussed by illustrative examples.

Fuzzy logic controller (FLC) is renowned for its adaptability and intuitive decision-making capabilities in active suspension systems, which face challenges stemming from unpredictable disturbances and complex vehicle dynamics. In this study, we introduce a novel optimization approach termed WW-PSO, which merges particle swarm optimization (PSO) with water wave optimization (WWO), aiming to elevate the performance of an FLC-based active suspension system. WWO efficiently solves optimization problems by simulating natural water wave behaviors. The hybridization of PSO and WWO leverages their complementary exploration and exploitation capabilities, resulting in improved performance and robustness of the optimized controller. The performance of the proposed controller, which is augmented with a linear quadratic controller (LQR), is evaluated across three scenarios featuring different road profiles and compared against other recent optimization methods which include genetic algorithm, tent sparrow search algorithm (Tent-SSA), and ST-PS-SO which is a combination of PSO, sewing trainee-based optimization, and symbiotic organism search. Simulation results show that the proposed WW-PSO significantly improves integral time absolute error (ITAE) for both body and wheel displacements, overshoot/undershoot (OS/US), and settling time. Specifically, the proposed method achieves a 53.37% improvement in ITAE, a 56.44% reduction in OS/US, and a 13.09% decrease in settling time for body displacements. For wheel displacements, it achieves a 52.90% improvement in ITAE, a 48.72% reduction in OS/US, and a 14.15% decrease in settling time. These enhancements demonstrate the hybrid method’s effectiveness in improving vehicle stability and passenger comfort across a range of road conditions.

Visual odometry is pivotal in robotics and autonomous driving, serving as a key component of visual simultaneous localization and mapping technology. In real-world scenarios, humans in local low-light conditions perceive less information, which can impact our judgments and actions. Similarly, visual odometry can become confused under these conditions, leading to compromised performance. To address the challenges posed by local low-light images on monocular visual odometry, we propose an unsupervised framework for monocular visual odometry. To the best of our knowledge, this is the first instance of unsupervised monocular visual odometry and local low-light image enhancement accomplished within a unified framework. Initially, we employ retinex theory and the discrete Fourier transform to decompose, filter, and synthesize the original image. For the filtering process, we propose a novel learnable global filtering network. Subsequently, we input the enhanced images into the depth and pose networks, generating the corresponding depth maps and inter-frame poses. Ultimately, we construct a photometric consistency loss, a depth loss, and a novel low-light smoothness loss to train the entire network. Through experimental validation, our method exhibits superior performance on the KITTI dataset. Furthermore, it demonstrates satisfactory generalization ability in unseen environments from the Oxford RobotCar dataset.

This study focuses on detecting and defending against false data injection attacks (FDIAs) on cyber-physical systems (CPSs). Firstly, recognizing the stealthy nature of FDIAs, deep reinforcement learning (DRL) is employed to design an automatic FDIA detector capable of learning different attack patterns. To enhance the robustness of the DRL algorithm, a new detection approach based on the improved proximal policy optimization (PPO) algorithm is devised to adapt to various FDIA modes. Secondly, to counteract the impact of FDIAs, an event-triggered model predictive control (MPC) approach is proposed to ensure the system swiftly returns to a stable state after being subjected to FDIAs. Lastly, the effectiveness of the proposed attack detector based on the DRL algorithm and the event-triggered model predictive controller is validated through a simulation example.

In this work, we demonstrate the efficiency of approximating nonlinear model predictive control (NMPC) using deep neural networks (DNN). We design an implicit NMPC for forward and backward motions of the truck trailer (TT) to handle complexity of nonlinear system dynamics. However, the high computational load of implicit MPC poses challenges for real-time implementation. To address this issue, we employ a DNN-based NMPC approximation to estimate parametric functions. As a result, the DNN-based NMPC approximation can mimic the optimal control policy of implicit MPC. Additionally, the average computation times for implicit NMPC and the DNN-based NMPC approximation in hardware-in-the-loop (HIL) tests are 36.541 ms and 0.031 ms, respectively.

Concentrated on the off-policy reinforcement learning method, this paper explores a model-free algorithm for addressing the robust fault-tolerant tracking problem in discrete-time linear systems with time-varying actuator faults and model uncertainties. Specifically, to determine the feedback control input, a dynamic optimization approach is developed based on measured data rather than exact information from system dynamics. Subsequently, a static optimization approach is established using solutions from the preceding dynamic optimization problem to compute the feedforward control input. Finally, numerical simulations are conducted to illustrate the feasibility and efficiency of the proposed solution.

For a class of continuous-time nonlinear systems with input constraints, a novel event triggered control (ETC) of integral reinforcement learning (IRL) based on sliding mode (SM) is proposed in this paper. Firstly, a SM surface-based performance index function is designed and the Hamiltonian equation is solved by the policy iteration algorithm. Secondly, the IRL technique is utilized to obtain the integral Bellman equation, which makes the controller do not need to know the drift dynamics. Thirdly, the ETC is introduced to reduce the communication burden and a triggering condition is designed to ensure the asymptotic stability of the system. Then, a critic neural network (NN) is used to learn the optimal value function to obtain the optimal tracking controller. Finally, the asymptotic stability of the whole closed-loop system and uniformly ultimately bounded of the critic NN weights are proved based on the Lyapunov theory. Simulation and comparison results demonstrate the effectiveness of the proposed method.

License plate recognition is an active research field as demands sharply increase with the development of Intelligent Transportation System (ITS). However, since the license plate recognition(LPR) is sensitive to the conditions of the surrounding environment such as a complicated background in the image, viewing angle and illumination change, it is still difficult to correctly recognize letters and digits on LPR. This study applies Deep Convolutional Neural Network (DCNN) to the license plate recognition. The DCNN is a method of which the performance has recently been proven to have an excellent generalization error rate in the field of image recognition. The proposed layer structure of the DCNN used in this study consists of a combination of a layer for judging the existence of a license plate and a layer for recognizing digits and characters. This learning method is based on Multi- Task Learning (MTL). Through experiments using real images, this study shows that this layer structure classifies digits and characters more accurately than the DCNN using a conventional layer does. We also use artificial images generated directly for training model.