The latest development of multi-robot collaborative systems has put forward higher requirements for multi-sensor fusion localization. Current position methods mainly focus on the fusion of the carrier’s own sensor information, and how to fully utilize the information of multiple robots to achieve high-precision positioning is a major challenge. However, due to the comprehensive impact of factors such as poor performance, variety, complex calculations, and accumulation of environmental errors used by commercial robots, the difficulty of high-precision collaborative positioning is further exacerbated. To address this challenge, we propose a low-cost and robust multi-sensor data fusion scheme for heterogeneous multi-robot collaborative navigation in indoor environments, which integrates data from inertial measurement units (IMUs), laser rangefinders, cameras, and so on, into heterogeneous multi-robot navigation. Based on Discrete Kalman Filter (DKF) and Extended Kalman Filter (EKF) principles, a three-step joint filtering model is used to improve the state estimation and the visual data are processed using the YOLO deep learning target detection algorithm before updating the integrated filter. The proposed integration is tested at multiple levels in an open indoor environment following various formation paths. The results show that the three-dimensional root mean square error (RMSE) of indoor cooperative localization is 11.3 mm, the maximum error is less than 21.4 mm, and the motion error in occluded environments is suppressed. The proposed fusion scheme is able to satisfy the localization accuracy requirements for efficient and coordinated motion of autonomous mobile robots.

The current situation of frequent small-scale accidents shows that the existing methods have not completely solved the problem of bearing failures, and new research methods need to be used to complete the study of bearing failures. To prevent the failure of rolling bearings and meet the need for timely detection of faults, this research is based on deep learning. Using the combination of deep transfer learning and metric learning methods, the identification and analysis of bearing multi-state vibration signals under different working conditions are carried out. The combination of SSAE-based similarity measurement criteria and deep transfer learning can reduce the differences between different domains. It is difficult to distinguish the data samples at the boundary and diagnose the problems that the physical meaning is difficult to understand. Through the bearing fault diagnosis analysis, the validity of the deep learning diagnosis model proposed in this paper is verified. The results show that the detection accuracy of the rolling bearing fault detection method based on LCM-SSAE is 0.6 percentage points higher than that of the rolling bearing fault detection method based on SSAE, which proves that the method is suitable for the fault detection of rolling bearing, and it also shows the effectiveness and robustness of the fault detection system of rolling bearing.

At present, deep learning-based intelligent detection technologies are progressively being applied to the field of steel surface defect inspection. An efficient defect inspection algorithm, YOLOv8n-GSE, is proposed to overcome low detection accuracy. We integrate a Global Attention Mechanism (GAM) into the YOLOv8n-based model’s backbone network to construct GAM attention layers, specifically designed to enhance feature representation for steel surface defects. We implement a CSP-ABAN structure in both the backbone and neck networks to improve the YOLOv8n-GSE model’s capacity for extracting steel surface defect features, while the ESCD head is introduced to process multi-scale features and capture defect information across various dimensions. By replacing standard convolutions with GSConv, the model achieves a balance between complexity and detection efficiency. Furthermore, the PIoUv2 loss function is adopted to address the limitations of CIoU. The experimental results demonstrate that the YOLOv8n-GSE model not only reduces computational complexity by 31% (to 5:6 GFLOPs) but also achieves a mean average precision of 94.2%, which represents a 5.4 percentage point improvement over the baseline YOLOv8n model. Compared with other object detection algorithms, YOLOv8n-GSE significantly enhances defect detection accuracy along with statistically significant reduction in miss rates.