Monitoring the assembly process is a challenge in the manual assembly of mass customization production, in which the operator needs to change the assembly process according to different products. If an assembly error is not immediately detected during the assembly process of a product, it may lead to errors and loss of time and money in the subsequent assembly process, and will affect product quality. To monitor assembly process, this paper explored two methods: recognizing assembly action and recognizing parts from complicated assembled products. In assembly action recognition, an improved three-dimensional convolutional neural network (3D CNN) model with batch normalization is proposed to detect a missing assembly action. In parts recognition, a fully convolutional network (FCN) is employed to segment, recognize different parts from complicated assembled products to check the assembly sequence for missing or misaligned parts. An assembly actions data set and an assembly segmentation data set are created. The experimental results of assembly action recognition show that the 3D CNN model with batch normalization reduces computational complexity, improves training speed and speeds up the convergence of the model, while maintaining accuracy. Experimental results of FCN show that FCN-2S provides a higher pixel recognition accuracy than other FCNs.

Article Highlights (1) A lightweight cross-scale feature fusion module is introduced to improve the feature fusion capability of the model. (2) The content-aware reassembly of feature module is introduced to enhance the quality of feature reconstruction. (3) Experiments are conducted with real datasets to verify the advantages of the improved method.

Optical sensor systems are essential for space target detection. However, previous studies have prioritized detection accuracy over model efficiency, limiting their deployment on resource-constrained sensors. To address this issue, we propose the lightweight space target real-time detection transformer (STar-DETR), which achieves a balance between model efficiency and detection accuracy. First, the improved MobileNetv4 (IMNv4) backbone network is developed to significantly reduce the model’s parameters and computational complexity. Second, group shuffle convolution (GSConv) is incorporated into the efficient hybrid encoder, which reduces convolution parameters while facilitating information exchange between channels. Subsequently, the dynamic depthwise shuffle transformer (DDST) feature fusion module is introduced to emphasize the trajectory formed by space target exposure. Finally, the minimum points distance scylla intersection over union (MPDSIoU) loss function is developed to enhance regression accuracy and expedite model convergence. A space target dataset is constructed, integrating offline and online data augmentation techniques to improve robustness under diverse sensing conditions. The proposed STar-DETR model achieves an $A{P}_{0.5:0.95}$ of 89.9%, successfully detecting dim and discontinuous streak space targets. Its parameter count and computational complexity are reduced by 64.8% and 41.8%, respectively, highlighting its lightweight design and providing a valuable reference for space target detection in resource-constrained optical sensors.

Augmented reality (AR) technology has been applied to the industrial field; however, at present, most industrial AR applications are developed for specific application scenarios, which are difficult to develop and have a long development cycle, and lack universality. To this end, this paper proposed a quick development toolkit for augmented reality visualization (QDARV) of a factory by using the script configuration and parsing approach. With QDARV, designers can quickly develop industrial AR applications, achieve AR registration based on quick response (QR) code and simultaneous localization and mapping (SLAM), and superpose information such as disassembly animations, industrial instruments, pictures, and texts on real scenes. In QDARV, an AR registration method based on SLAM map reloading is proposed. Through saving and reloading the map and the configuration and analysis of the SLAM-AR display content script, the AR scene configured by the designer is displayed. Meanwhile, the objects detected using YOLOv5 are used as a landmark to assist the SLAM system in positioning. The QDARV can be applied to realize AR visualization of factories with a large area.

The soft gripper has received extensive attention, due to its good adaptability and flexibility. The dielectric elastomer (DE) actuator as a flexible electroactive polymer that provides a new approach for soft grippers. However, they have the disadvantage of having a poor rigidity. Therefore, the optimization design method of a rigid-flexible soft finger is presented to improve the rigidity of the soft finger. We analyzed the interaction of the rigid and soft materials, using the finite element method (FEM), and researched the influence of the parameters (compression of the spring and pre-stretching ratio of the DE) on the bending angle. The optimal parameters were obtained using the FEM. We experimentally verified the accuracy of the proposed method. The maximum bending angle is 19.66°. Compared with the theoretical result, the maximum error is 3.84%. Simultaneously, the soft gripper with three fingers can grasp various objects and the maximum grasping quality is 11.21 g.

In the Industrial Internet of Things (IIoT) field, the diversity of devices and protocols leads to interconnection challenges. Conventional Modbus Transmission Control Protocol (TCP) to Remote Terminal Unit (RTU) gateways suffer from high overhead and latency of the TCP protocol stack. To enhance real-time communication while ensuring reliability, this study applies Markov chain theory to analyze User Datagram Protocol (UDP) transmission characteristics. An Advanced UDP (AUDP) protocol is proposed by integrating a Cyclic Redundancy Check (CRC) check mechanism, retransmission mechanism, Transaction ID matching mechanism, and exponential backoff mechanism at the UDP application layer. Based on AUDP, a Modbus AUDP-RTU gateway is designed with a lightweight architecture to achieve bidirectional conversion between Modbus AUDP and Modbus RTU. Experimental validation and Markov chain-based modeling demonstrate that the proposed gateway significantly reduces communication latency compared to Modbus TCP-RTU and exhibits higher reliability than Modbus UDP-RTU.