With the rapid development of autonomous driving technology, the demand for multi-sensor fusion in environmental perception systems is increasing. Four-dimensional (4D) millimeter-wave radar has become one of the critical sensors in autonomous driving due to its stable performance under complex weather and lighting conditions. Although 4D millimeter-wave radar improves object detection accuracy by adding elevation angle information and increasing point cloud density, its sparse point clouds and noise issues limit its independent application. Therefore, the fusion of 4D millimeter-wave radar with vision sensors has become key to enhancing perception accuracy in autonomous driving. However, traditional extrinsic calibration methods rely on cumbersome manual operations, making it challenging to meet the requirements for efficient automated calibration. To address this issue, this study proposed an automated extrinsic calibration method for 4D millimeter-wave radar and visual images based on a calibration board. The method first designs a calibration board with ChArUco markers, red circular rings, and corner reflectors, and then automatically extracts image coordinates and radar point cloud coordinates of the calibration points using a circle detection algorithm and a corner reflector detection algorithm. Furthermore, a method for calibration data acquisition and validation was proposed using simulations in 3D Max and Unity. Finally, the performance of direct linear transformation (DLT) and extrinsic calibration (EC) methods is compared through experiments to evaluate calibration accuracy. Experimental results indicate that the designed calibration board and automated calibration algorithm effectively reduce manual operations, and the EC method demonstrates higher calibration stability and accuracy when more calibration points are involved.

Rolling bearings, as a type of precision mechanical component, are widely used in modern industrial machinery and equipment. It is of great significance to diagnose bearing faults using reasonable methods during bea-ring operation. However, in the actual complex and ever-changing environment, the collection of vibration signals often faces challenges such as limited sample sizes, noise interference, and operating condition variations, resulting in low fault diagnosis accuracy. To address the problem of small-sample rolling bearing fault diagnosis under noise interference and variable operating conditions, this paper proposed a meta-learning denoising model based on prototype domain enhancement (Meta-DAE). Firstly, a small-sample fault dataset based on time-frequency diagrams was constructed, and a deep convolutional generative adversarial network was introduced for data preprocessing to gene-rate a pseudo-sample set with a similar distribution. Then, the fault sample set was input into Meta-DAE for adaptive feature extraction. Meta-DAE adopts a prototype domain enhancement strategy to make prototype points of the same category more closely clustered in the embedding space. At the same time, an encoder with noise reduction performance was constructed, and a target function based on prototype domain enhancement and denoising was designed. By fine-tuning the model under small-sample conditions, the noise robustness and classification accuracy of the model were improved. Experimental results of small-sample fault diagnosis under noise interference and variable operating conditions show that, compared to other models, the proposed model demonstrates strong noise robustness. Under -8 dB strong noise interference, the model achieves a classification accuracy improvement of 35.78% to 57.25% using only 10 samples for fine-tuning.

The development of large-scale communication systems puts higher requirements on traditional filter design. Sparse finite impulse response (FIR) filters have the characteristics of low computational complexity and low implementation cost, but conventional convex relaxation approximation design methods produce additional approximation errors, exhibit suboptimal sparsity, and involve complex solving processes. To address the issue of high implementation costs caused by the large number of multipliers in FIR filter design, this paper proposed a sparse FIR filter design method based on a weighted least squares criterion. Firstly, the norm of the initial sparse representation is replaced based on the properties of different norms, thereby improving the objective function. This modification maintains sparsity while addressing the challenge of directly solving non-convex functions. Next, the target problem was reformulated as the difference between two convex sub-problems. Simplified sub-problems were constructed according to iterative rules, and an alternating solution method was adopted to further enhance solving efficiency and reduce complexity. Finally, after determining the positions of zero coefficients, a weighted least squares problem was solved to further reduce approximation errors. The simulation results show that compared with the existing sparse filter solving methods, the proposed method can improve the coefficient sparsity performance of FIR filters, reduce the number of multipliers and obtain a compromise between root-mean-square error and maximum error in the case of sparsity enhancement. Meanwhile, the computational solving time is significantly reduced, and solving efficiency is notably improved.

Aiming at the trajectory tracking control problem of four-Mecanum-wheel AGV (Automatic Guided Vehicle), this study designs a controller named MPC+ELM-SMC, which connects model prediction control and adaptive sliding mode control to improve the control accuracy and stability, and to enhance the control process hierarchy, specificity and effectiveness. At the kinematic level, the trajectory tracking error model is established, which is transformed into a quadratic programming problem, and constraints are added to solve the optimal solution of the quadratic programming online with the rolling optimization of the model predictive control, and the AGV pose error is converted into the expected output of the wheel speed. At the dynamic level, sliding mode control is used to obtain the wheel torque output and to realize the wheel’s tracking of the expected speed. The ELM (Extreme Learning Machine) neural network with fast and accurate approximation ability is introduced to carry out online observation of the model uncertainty and unknown interference, and the adaptive interference is offset in combination with sliding mode control to further improve the robustness of the controller. Under the conditions of cosine disturbance and pulse interference, the controller is simulated and verified. Compared with PID control, the MPC+SMC cascade controller has obvious advantages in tracking effect. Moreover, compared with the cascade controller observed by RBF (Radial Basis Function) neural network, the ELM observer is more robust to interference, for instance, the observation effect remains above 95% under various rotational speed conditions, the tracking error of the proposed control method is one order of magnitude smaller than other methods in multiple indicators, with a maximum deviation of only millimeters. Finally, an experimental platform is set up to perform actual trajectory tracking experiment, and the results verify the practicability and feasibility of the proposed controller.

In the industrial Internet of Things, the reliability of mobile edge computing largely depends on the wireless channel conditions. In order to process the influence of imperfect channel state information to the system, this paper proposed a digital twin assisted mobile edge computing energy consumption optimization method. For the task offloading problem in industrial Internet of Things, a digital twin model of devices and channels in the edge computing system was established. Considering imperfect channel state information, the joint optimization of offloading decisions, transmission power, channel resources, and computational resources is performed with the aim of minimizing the total system energy consumption. To deal with the proposed nonlinear non convex problem of mixed integers, the probabilistic delay constraint was transformed and the original problem was decomposed into two sub-problems, and a joint optimization algorithm with the assistance of digital twins based on continuous convex approximation was proposed. Firstly, the original problem was relaxed to obtain resource allocation schemes and task offloading priorities. Then, the offloading priority of each terminal device was sorted in descending order. The complete task offloading scheme was obtained by solving the iterative optimization problem. Finally, simulation results show that, compared to other benchmark schemes, the proposed computational offloading optimization scheme significantly reduces the total energy consumption of the system.

Metal-oxide thin-film transistor (MO-TFT) can be utilized to create a flexible wearable system for detec-ting heart rate signals. However, the lack of high-performance complementary devices in MO-TFTs results in low gain for the implemented preamplifiers. Additionally, the relatively poor performance of MO-TFT devices poses challenges for the design of subsequent modules. In order to improve the gain of the preamplifier and reduce the performance requirements of the subsequent digital circuit, this paper proposed a common source common gate capacitor bootstrap structure preamplifier. The preamplifier was mainly composed of an external coupling bias module and a core amplifier module. The core amplifier module uses a capacitor bootstrapping technique known for its excellent stability, large output voltage swing, and low power consumption. This technique is combined with a cascode structure to enhance the overall gain of the circuit. The external coupling bias module utilizes an AC-coupled external bias structure that features low power consumption, high input impedance, and straightforward operating point setting, thereby meeting the bandpass requirements of the preamplifier for heart rate signal detection. The proposed preamplifier was designed and fabricated using a 10 μm IZO-TFT process. The test results indicate that with a 20 V power supply voltage, the circuit has a gain of 35 dB, a bandwidth of 2 Hz~2 kHz, a noise of 118.2 μV, and a power consumption of 0.1 mW. The presented preamplifier meets the requirements for detecting heart rate signals. In comparison to the current MO-TFT heart rate signal detection preamplifier, the gain has been increased by about 10 dB, which reduces the performance requirements of the subsequent digital module on the device, and is beneficial to achieve the digitalization of analog signals and maintain the signal integrity.

To address the problems of high operational energy consumption and low efficiency in air-conditioning system for electric drive workshops, this paper proposed a dynamic energy-saving optimization method based on the IBK-IPS algorithm, taking into account the mutual constraints of each equipment of the air-conditioning system. Firstly, the influence mechanism among system components were analyzed, and mathematical models of energy consumption and constraints conditions for each device were established to an objective function for system energy consumption. Then, an improved Black Kite-Particle Swarm (IBK-IPS) algorithm was introduced to optimize operational parameters such as water temperature, flow rate, and air volume, thereby improving the accuracy and effectiveness of system parameter control. Subsequently, a simulation model of the cooling water system and chilled water system of the air-conditioning system is developed using the Simulink platform to evaluate the performance and accuracy of the parameter optimization. Finally, the method is practically applied in an electric drive workshop to verify the practical effect and feasibility of the proposed method. The results of simulation experiments and practical application tests show that: the operational energy consumption of the system is effectively reduced, achieving an energy-saving rate of 11.23%~34.68%, and improves operational energy efficiency by 11.53%~41.75%. Compared with the PS, BK, and BK-PS algorithms, the IBK-IPS algorithm delivers superior energy-saving performance, with convergence speeds improved by 27.27%, 61.90%, and 69.23%, respectively. In real-world testing under five different load conditions, the optimized system achieved energy-saving rates of 22.61%, 17.24%, 7.48%, 14.97%, and 12.64%, respectively. In summary, the energy-saving optimization method proposed in this paper can effectively solve the problem of high energy consumption and low efficiency of the air-conditioning system operation in the electric drive workshop, which has good energy-saving effect and practicability, and can provide new ideas for the research of energy-saving optimization of air-conditioning system.

To address the issues of low classification accuracy and poor generalization in existing acoustic scene classification methods, this paper proposed a novel acoustic scene classification method based on reducing high-frequency reverberation and a frequency-domain residual shrinkage network with multi-scale attention, named RF-DRSN-EMA. Firstly, according to the principle of reducing sound reverberation, this paper introduced a redu-cing high-frequency reverberation method. This method attenuated only the high-frequency reverberation while preserving essential frequency information in other bands. As a result, speech intelligibility was enhanced, and the impact of speech distortion was minimized. Secondly, based on the deep residual shrinkage network, the proposed RF-DRSN-EMA integrates an improved frequency-domain self-calibration mechanism and a multi-scale attention module. The network used RF self-calibration module with a long-short residual structure to mitigate feature collapse, enabling efficient extraction of frequency-domain information. A multi-scale attention module was then applied at the output of each unit to highlight relevant information, further enhancing the model’s representation capacity. Finally, the proposed method is evaluated on three benchmark datasets: ESC-10, UrbanSound8K, and DCASE2020 Task 1A. The results show that the proposed high-frequency reverberation reduction method effectively suppresses high-frequency reverberation and background noise while eliminating redundant features, resulting in minimal speech quality degradation. The RF-DRSN-EMA network achieves efficient frequency-domain denoising and feature extraction, reaching classification accuracies of 98.00%, 93.42%, and 72.80% on the three datasets, respectively. These results confirm the effectiveness and generalizability of the proposed method.

Feeding-assistance robots are key equipment in promoting the modernization and transformation of animal husbandry. The rapid and accurate identification of feeding targets is essential for enabling intelligent feed-pushing, while balancing segmentation accuracy and operational efficiency is crucial for ensuring the overall performance of recognition algorithms—an important topic in the field of intelligent livestock management. To address the mismatch between segmentation accuracy and processing efficiency in current dairy cow feeding target recognition methods, this paper proposed a real-time feeding target recognition method (RTFTR) based on an optimized Segment Anything Model (SAM). Built on the SAM-det architecture, RTFTR first introduces lightweight image encoder and object detector, along with a parallelized buffer queue design, to balance the operational efficiency of each module and enhance inference speed. It then employs a High-Quality (HQ) token mechanism to enhance the feature space decoding capacity, optimizes the mask decoder, and applies stage-wise training tailored to feeding targets to improve segmentation accuracy. Experimental results show that the proposed method ensures inference efficiency while enhancing segmentation accuracy. In the task of cow feeding target recognition, the method achieves a segmentation accuracy of 98.7% for cows, 96.4% for feed, 99.2% for bunk, with an overall average accuracy of 98.1%, and a processing speed of 52.9 f/s, meeting the application requirements for cow feeding target recognition in complex environments and limited robotic computational resources.