Most face recognition methods rely on deep convolutional neural networks (CNNs) that construct multiple layers of processing units in a cascaded form and employ convolution operations to fuse local features. However, these are not conducive modeling the global semantic information lack attention important facial feature regions their spatial relationships. In this work, Group Depth-Wise Transpose Attention (GDTA) block is designed effectively capture both representations, mitigate issue...

Abstract As an important energy storage device, lithium-ion batteries have vast applications in daily production and life. Therefore, the remaining useful life (RUL) prediction of such is great significance, which can maintain efficacy reliability system powered by batteries. For predicting accurately, adaptive hybrid battery model improved particle filter (PF) are developed. First, constructed, a combination empirical long short-term memory (LSTM) neural network that it could characterize...

In this article, we propose a gradient-based event-driven model predictive control (GEMPC) algorithm with state-dependent threshold for nonlinear systems additive disturbances and input state constraints. Firstly, novel strategy is constructed in the light of error gradient between optimal prediction real one, which could ensure Zeno-free property via positive triggering interval. Subsequently, mechanism dual-mode are combined to establish GEMPC framework, further reduce computing burden...

Abstract A quasi‐differential type event‐triggered model predictive control (ET‐MPC) framework for continuous‐time linear systems with additional disturbances is constructed. Different from the existing ET‐MPC, triggering condition of proposed method focused on differences errors between actual states and best prediction sequence at two consecutive sampling moments. Its advantage that dynamic characteristics state changes can be better considered, which will achieve a more effective balance...

This paper attempts to replace the traditional manual slag offloading of magnesium smelting in Pidgeon process with robotic removal.Specifically, high-temperature infrared dot matrix was used measure positions indirectly; faster region-based convolutional neural network (Faster R-CNN) trained thermal image reduction jar as dataset; isothermal plotted based on centers, and adopted detect opening direction positions.The indirect measurement results show that actual internal temperature can be...

Aiming at solving the control problem of a constrained and perturbed underwater robot, method was proposed by combining self-triggered mechanism nonlinear model predictive (NMPC). The theoretical properties kinematic as well corresponding MPC controller, are first studied. Then, novel technique for determining next update moment both optimal system state is developed. It further rigorously proved that algorithm can (1) stabilize closed-loop robot system, (2) reduce time (3) save information...

Traditional YOLO models face a dilemma when it comes to dim detection targets: the accuracy increases while speed inevitably reduces, or vice versa. To resolve this issue, we propose novel DME-YOLO model, which is characterized by establishment of backbone based on YOLOv7 and Dense blocks. Moreover, through application feature multiplexing, both parameters floating-point computation were decreased; therefore, defect process was accelerated. We also designed multi-source attention mechanism...

Abstract In this paper, a model predictive control (MPC) algorithm framework based on machine learning (ML) is proposed for mobile robot system, which integrates event-triggered mechanism (ETM) and parameters self-tuning (PSM). Firstly, the kinematic of established, MPC-based path tracking controller designed. Secondly, PSM MPC designed kernel extreme optimized by sparrow search (SSA-KELM). Then, two mechanisms are to determine whether perform optimization solution tuning, respectively....

Nonlinear progressive failure study of surrounding rock is important for the stability analysis underground engineering projects. Taking a deep-buried tunnel in Chongqing as an example, three dimensional(3-D) physical model was established based on similarity theory. To satisfy requirement physical–mechanical properties, such elastic modulus, compressive strength and Poisson ratio, materials were developed. Using full inner-spy photograph technology, deformation process studied under...

In this paper, we propose a self-balancing mobile robot system and further design control method to improve the resistance external impact adaptability continuous loads when is at fixed point. Firstly, put forward in paper compared with classic two-wheeled show its structural advantages. Secondly, mechanical part electrical of are introduced, mathematically modeled analyzed. Finally, an adaptive double-fuzzy anti-integral saturation PID developed dual-loop on proposed robot. Experiments that...

A collaborative evolutionary optimization method for the layout of hydraulic manifold blocks was presented according to its structure feature, an design mathematic model in spatial established, and it solved based on ant colony algorithm.Special mention given here demonstrate improvement current multiple algorithms transfer policy rules heuristic information which made Cooperative arrangement hole routing.Finally, simulation carried out with MATLAB verify feasibility improved algorithm.The...

Abstract This study examined the control problem of continuous‐time nonlinear systems under a sample‐and‐hold structure. A new integral‐type double self‐triggered (ST) model predictive (MPC) approach is developed by combining novel ST strategy with MPC, which significantly decreases sampling frequency system, saving computation/communication resources. In contrast to traditional strategy, two self‐triggering mechanisms are designed based on mode state and integral difference between optimal...

Stress concentration damage is one of the most troublesome phenomena encountered in practical engineering structure. In order to reduce damage, it very important for stress factors determine arbitrarily and accurately. At present, more attention has been given problem concentration, modern structural reliability theory neural network technique have improved rapidly. But difficult be solved case concentration. Neural (NN) used this work simulate relationship between basic random variables...

Due to the popularity of 5G connectivity and The Internet Things sensors, deep learning algorithms are being extended edge devices. Compared with AI(Artificial Intelligence) cloud platforms, deployment neural networks on devices must focus low power consumption, latency, stability reliability. In recent years, development lightweight network architecture has provided a basis for However, shortcomings networks, such as overconfidence, vulnerability adversarial attack, easy over fitting when...