Robot control based on visual information perception is a hot topic in the industrial robot domain and makes robots capable of doing more things in a complex environment. However, complex visual background in an industrial environment brings great difficulties in recognizing the target image, especially when a target is small or far from the sensor. Therefore, target recognition is the first problem that should be addressed in a visual servo system. This paper considers common complex constraints in industrial environments and proposes a You Only Look Once Version 2 Region of Interest (YOLO-v2-ROI) neural network image processing algorithm based on machine learning. The proposed algorithm combines the advantages of YOLO (You Only Look Once) rapid detection with effective identification of ROI (Region of Interest) pooling structure, which can quickly locate and identify different objects in different fields of view. This method can also lead the robot vision system to recognize and classify a target object automatically, improve robot vision system efficiency, avoid blind movement, and reduce the calculation load. The proposed algorithm is verified by experiments. The experimental result shows that the learning algorithm constructed in this paper has real-time image-detection speed and demonstrates strong adaptability and recognition ability when processing images with complex backgrounds, such as different backgrounds, lighting, or perspectives. In addition, this algorithm can also effectively identify and locate visual targets, which improves the environmental adaptability of a visual servo system

With the rapid advancements in surface mount technology (SMT), microelectronic packaging products are becoming smaller, denser and defect-free. This has led to stricter requirements for surface quality inspection. In the realm of industrial visual inspection, there is a growing demand for three-dimensional (3D) imaging technology to achieve high-precision and fast image acquisition, as well as for AI-based online inspection models to enhance real-time defect detection. This paper primarily focuses on first- and second-level package products for SMT microelectronics and discusses the main optical imaging technologies, analyzing the principles and limitations. Next, this paper analyzes the current research status of intelligent detection models based on deep learning in industrial environments and discusses current technical issues in data augmentation, small object learning, and model lightweight. The paper also summarizes the characteristics and applications of the latest representative products, comparatively analyzes the gap between developed and developing countries. Finally, this paper outlines four major challenges and future research trends in this field, aiming to promote the intelligent evolution of related visual inspection equipment.

Recently, flexible airflow sensors have attracted significant attention due to their impressive characteristics and capabilities for airflow sensing. However, the development of high‐performance flexible airflow sensors capable of sensing airflow over large areas remains a challenge. In this work, it is proposed that a hair‐like flexible airflow sensor, based on laser direct writing and electrostatic flocking, offers an efficient technology for airflow sensing. The airflow sensor exhibits high sensitivity (5.25% s m ⁻¹ ), fast response time (39.83 ms), a wide detection range (3.48 – 18.36 m s ⁻¹ ), minimal disturbance to the airflow field, simplified fabrication, and cost efficiency. The airflow sensor, with excellent conformal monitoring capabilities, is capable to detect airflow on different surfaces. The sensor's excellent flexibility and efficient fabrication process enable it to be easily integrated into arrays and deployed on large‐area surfaces to provide information, including airflow velocity, direction, and point of incidence. The hair‐like airflow sensor has significant potential for applications in environmental monitoring, intelligent robots, wearable electronics, and tactile sensing.

Gas-insulated switchgear (GIS) serves as the core equipment in substations. Its partial discharge detection requires ultrasonic sensors to be precisely aligned with millimeter-level measurement points. However, existing technologies face three major bottlenecks: the lack of surface texture on GIS makes visual feature extraction difficult; strong electromagnetic interference in substations causes image noise and loss of feature point tracking; and fixed gain control easily leads to end-effector jitter, reducing positioning accuracy. To address these challenges, this paper first employs AprilTag visual markers to define GIS measurement point features, establishing an image-based visual servo model that integrates GIS surface curvature constraints. Second, it proposes an adaptive gain algorithm based on model predictive control, dynamically adjusting gain in real-time according to visual error, electromagnetic interference intensity, and contact force feedback, balancing convergence speed and motion stability. Finally, experiments conducted on a GIS inspection platform built using a Franka Panda robotic arm demonstrate that the proposed algorithm reduces positioning errors, increases positioning speed, and improves positioning accuracy compared to fixed-gain algorithms, providing technical support for the engineering application of GIS partial discharge detection robots.

A smart manufacturing prototype called iCandy Box, used for hybrid packing of assorted candies, was designed to study and verify the cyber-physical control methods. The prototype is aimed to provide personalized consumption, as it can perform flexible and customized production. The prototype is powered by a cloud-edge-end enabled collaborative information framework, which can support both industrial big data and artificial intelligence applications. Furthermore, it is characterized by modularization and interdisciplinarity; therefore, it can be used to carry out both experiments and training in several major fields, including smart manufacturing and IoT. The experimental results have shown that the prototype can carry out hybrid production, paving the way for the study and verification of cyber-physical control methods.

Robot program interpreter is an important tool for robot to realize real-time control and fast feedback. In order to simplify the process of robot programming and improve the interpretation efficiency of robot program, an easy-to-use and efficient industrial robot program interpreter is designed and implemented in this paper. The interpreter divides the robot program into two parts: variable definition and instruction call. The process of interpretation includes lexical analysis, syntactic analysis, semantic analysis and instruction interpretation modules. Using flex and bison tools to assist in the generation of lexical and syntactic analysis programs, this paper proposes child-sibling notation (CSN) to construct a syntax tree. In semantic analysis, the red-black tree structure of the map container is used to create a symbol table and record variable information. By the way of presetting type checking codes, errors in the program can be reported and handled. Finally, the interpreter traverses the syntax tree with depth-first algorithm, and calls the corresponding control function while interpreting the instruction sentence to execute the motion of robot. The experimental results show that the designed interpreter has high efficiency and stability in interpreting the robot program and meets the operational requirements of industrial robots.