Many kinds of peg-in-hole assembly strategies for an industrial robot have been reported in recent years. Most of these strategies are realized by utilizing visual and force sensors to assist robots. However, complex control algorithms that are based on visual and force sensors will reduce the assembly efficiency of a robot. This issue is thoughtless in traditional assembly strategies but is critical to further improve the efficiency of assembly automation. In this work, a new assembly strategy that is based on a displacement sensor and a variable compliant center is proposed to improve robot performance in assembly tasks. First, an elastic displacement device for this assembly strategy is designed, and its performance is analyzed. The displacement signal generated by the displacement sensor is used to detect the contact state of the peg and hole and to guide the robot to adjust the posture. Second, an assembly strategy, including the advantages of passive compliance and active compliance, and a simple assembly control system are designed to improve the assembly efficiency. Last, the effectiveness of the proposed assembly method is experimentally verified using a robot with 6 degrees of freedom and a chamferless peg and hole with a small clearance (0.1 mm). The experimental results show that the assembly strategy can successfully complete the precision peg-in-hole assembly and assist the robot in accurate assembly in industrial applications.

Image segmentation is one of the most important research directions in computer fields and is also an important component of pattern recognition. In RoboCup competitions, the accurate positioning of the target object is particularly important to the control of the robots. An image segmentation method based on edge and color features is proposed in the present paper. Firstly, the image pixel color features are extracted in multi-color space, and the edge features of the objects are extracted by an adaptive Canny operator. The color features set the entropy, mean, and variance of the pixels in the component image. Then the color and edge features of the image are clustered by Fuzzy C-means. The clustering result of FCM is used as the input of the RBF-SVM model. Finally, the training model is applied to segment the image and output the position of the target object. The present experiment first verified the performance of the multi-color space model by simulation experiments, then conducted image segmentation experiments in images from RoboCup competition, and used Region Circularity and Boundary Center Deviation as evaluation indicators to analyze the performance of the algorithm. The experimental results showed that this method has better performance in image segmentation. Compared with existing methods, it reduced the segmentation time and improved the image segmentation performance.

With the prevailing application of new materials and the higher requirements for the quality and efficiency of production in the equipment manufacturing industry, traditional assembly methods can hardly meet the needs of large-scale production, especially in the field of high-precision assembly. Robot assembly guided by visual perception has become the key of the research in the field of engineering technology. It requires higher accuracy of robot visual perception and the control over force, position and so on. However, in 3C assembly, most products are made of transparent materials such as glass. Because of the transparency and specular reflection of the surface, 3D reconstruction of transparent objects is a very difficult problem in computer vision, in that the traditional visual perception methods could not be accurate enough. The present research proposes a bionic active sensing algorithm for 3D perception and reconstruction and realizes high-precision 3D by applying the registration algorithm. The purpose is to solve the problems existing in the traditional visual perception method, such as difficulties in achieving active sensing, low accuracy of point clouds registration, and complex computation. The results of the experiments show that the present method is efficient and accurate in 3D reconstruction. It reduces the planar reconstruction error to 0.064 mm and the surface reconstruction error to 0.177 mm.

In view of the problem that the highway network measurement system cannot accurately evaluate the load-bearing quality of electric power business, a quantitative evaluation method of 5G network quality for electric power business is proposed, and a four-level composite weighted evaluation model of 5G network bearing quality of electric power business is constructed. The wireless network quality evaluation index system is established, which covers the key performance indicators of all kinds of electric power business. The quantitative evaluation method of service bearing quality is proposed and realized systematically. The system supports power business simulation, terminal signal quality, end-to-end delay, reliability and other test capabilities. It has the functions of power multi-service carrying quality evaluation, comparison of 5G multi-network test results, and test analysis of key indicators of 5G network. It can accurately judge the network quality of the business side, scientifically evaluate and verify the service quality of 5G network carrying power business, make up for the shortcomings of the traditional road-oriented network test system, and quickly locate the problem, which plays an important role in the operation, maintenance and optimization of power 5G network slices.

At present, the semantic information segmentation algorithms mainly include FCN (Fully Convolutional Network), PSPNet (Pyramid Scene Parsing Network), Deeplab and so on. In view of the inadequate results of features extracted by these algorithms from RGB image, a hybrid fully convolutional autoencoder neural network (HFCAN) structure, which introduces fully convolutional neural network and stacked sparse autoencoder, is proposed in this paper. Using the FCN to generate the thermal high-dimensional feature map of the shelf commodity, and then performing the up-sampling operation on the segmented feature map. During the up-sampling operation, the convolution features are refined by the stacked sparse autoencoder (SAE), and the image boundary details are retained, so that the classification results are more accurate. The experimental results show that the hybrid fully convolutional autoencoder model proposed in this paper can not only shorten the training time and testing time of neural network by nearly 50%, but also improve the accuracy of shelf commodity identification by nearly 95%.