This paper introduces an enhanced Whale Optimization Algorithm, named the Multi-Swarm Improved Spiral Whale Optimization Algorithm (MISWOA), designed to address the shortcomings of the traditional Whale Optimization Algorithm (WOA) in terms of global search capability and convergence velocity. The MISWOA combines an adaptive nonlinear convergence factor with a variable gain compensation mechanism, adaptive weights, and an advanced spiral convergence strategy, resulting in a significant enhancement in the algorithm’s global search capability, convergence velocity, and precision. Moreover, MISWOA incorporates a multi-population mechanism, further bolstering the algorithm’s efficiency and robustness. Ultimately, an extensive validation of MISWOA through “simulation + experimentation” approaches has been conducted, demonstrating that MISWOA surpasses other algorithms and the Whale Optimization Algorithm (WOA) and its variants in terms of convergence accuracy and algorithmic efficiency. This validates the effectiveness of the improvement method and the exceptional performance of MISWOA, while also highlighting its substantial potential for application in practical engineering scenarios. This study not only presents an improved optimization algorithm but also constructs a systematic framework for analysis and research, offering novel insights for the comprehension and refinement of swarm intelligence algorithms.

Inspection robots, which improve hazard identification and enhance safety management, play a vital role in the examination of high-risk environments in many fields, such as power distribution, petrochemical, and new energy battery factories. Currently, the position precision of the robots is a major barrier to their broad application. Exact kinematic model and control system of the robots is required to improve their location accuracy during movement on the unstructured surfaces. By a virtual engine and digital twins, this study put forward a visualization monitoring and control system framework which can address the difficulties in the intelligent factories while managing a variety of data sources, such as virtual–real integration, real-time feedback, and other issues. To develop a more realistic dynamic model for the robots, we presented a neural-network-based compensation technique for the nonlinear dynamic model parameters of outdoor mobile robots. A physical prototype was applied in the experiments, and the results showed that the system is capable of controlling and monitoring outdoor mobile robots online with good visualization effects and high real-time performance. By boosting the positional accuracy of robots by 18% when navigating obstacles, the proposed precise kinematic model can increase the inspection efficiency of robots. The visualization monitoring and control system enables visual, digital, multi-method, and complete real-time inspections in high-risk factories, such as new energy battery factories, to ensure the safe and stable operations.

This paper addresses the challenge of detecting unknown or unforeseen obstacles in railway track transportation, proposing an innovative detection strategy that integrates an incremental clustering algorithm with lightweight segmentation techniques. In the detection phase, the paper innovatively employs the incremental clustering algorithm as a core method, combined with dilation and erosion theories, to expand the boundaries of point cloud clusters, merging adjacent point cloud elements into unified clusters. This method effectively identifies and connects spatially adjacent point cloud clusters while efficiently eliminating noise from target object point clouds, thereby achieving more precise recognition of unknown obstacles on the track. Furthermore, the effective integration of this algorithm with lightweight shared convolutional semantic segmentation algorithms enables accurate localization of obstacles. Experimental results using two combined public datasets demonstrate that the obstacle detection average recall rate of the proposed method reaches 90.3%, significantly enhancing system reliability. These findings indicate that the proposed detection strategy effectively improves the accuracy and real-time performance of obstacle recognition, thereby presenting important practical application value for ensuring the safe operation of railway tracks.

Automated guided vehicles (AGVs) have emerged as crucial machinery in enterprise production and transportation as intelligent factories have gained traction. This study proposes a multi-source data enhancement fusion convolutional neural network (ME-CNN) approach for the mechanical fault diagnosis of heavy-duty AGVs. Specifically, an improved stacked autoencoder (SAE) was employed to augment the available data and address the issue of insufficient data samples during fault diagnosis. Subsequently, the augmented data were processed using a continuous wavelet transform (CWT). This method employs a combination of diverse measurement signals and data sources to facilitate a more precise and resilient diagnosis of heavy-duty AGVs. In addition, to further promote the coverage of smart factories, this method is combined with digital mirroring, and the virtual platform receives data to monitor the status of heavy-loaded AGVs in real-time and determine the type of fault. Following testing, the system response time was less than 500 ms, allowing for the timely diagnosis of faults. Finally, the validity of the method was verified by comparing a factory-heavy load AGV with various other algorithms, resulting in an accuracy rate of 97.0% for fault diagnosis.

In the context of industrial robot maintenance and assembly, workers often suffer from work-related musculoskeletal disorders (WRMSDs). This paper proposes a multi-scale, multi-stage pose recognition method (MMARM-CNN) based on convolutional neural networks to provide ergonomic intervention. The method leverages computer vision technology to enable non-contact data acquisition, reducing the interference of physiological and psychological factors on assessment results. Built upon the baseline yolov8-pose framework, the method addresses complex maintenance environments, which are prone to occlusion, by introducing the Lightweight Shared Convolutional Detection Head-pose (LSCD-pose) module, Multi-Scale Channel Attention (MSCA) mechanism, and Efficient Multi-Scale Patch Convolution (EMSPC) module, enhancing the model’s feature extraction capabilities. The MMARM-CNN model was validated using the MS COCO 2017 dataset and robot assembly data collected under laboratory conditions. The experimental results show that the MMARM-CNN achieved an accuracy improvement, reaching 0.875 in the mAP@0.5 evaluation. Overall, this method demonstrates significant potential in advancing the automation and intelligence of ergonomic interventions.

In this paper, an Enhanced Greylag Goose Optimization Algorithm (EGGO) based on evolutionary game theory is presented to address the limitations of the traditional Greylag Goose Optimization Algorithm (GGO) in global search ability and convergence speed. By incorporating dynamic strategy adjustment from evolutionary game theory, EGGO improves global search efficiency and convergence speed. Furthermore, EGGO employs dynamic grouping, random mutation, and local search enhancement to boost efficiency and robustness. Experimental comparisons on standard test functions and the CEC 2022 benchmark suite show that EGGO outperforms other classic algorithms and variants in convergence precision and speed. Its effectiveness in practical optimization problems is also demonstrated through applications in engineering design, such as the design of tension/compression springs, gear trains, and three-bar trusses. EGGO offers a novel solution for optimization problems and provides a new theoretical foundation and research framework for swarm intelligence algorithms.