Research on Mobile Robot Navigation Method Based on Semantic Information.

This paper proposes a solution to the problem of mobile robot navigation and trajectory interpolation in dynamic environments with large scenes. The solution combines a semantic laser SLAM system that utilizes deep learning and a trajectory interpolation algorithm. The paper first introduces some open-source laser SLAM algorithms and then elaborates in detail on the general framework of the SLAM system used in this paper. Second, the concept of voxels is introduced into the occupation probability map to enhance the ability of local voxel maps to represent dynamic objects. Then, in this paper, we propose a PointNet++ point cloud semantic segmentation network combined with deep learning algorithms to extract deep features of dynamic point clouds in large scenes and output semantic information of points on static objects. A descriptor of the global environment is generated based on its semantic information. Closed-loop completion of global map optimization is performed to reduce cumulative error. Finally, T-trajectory interpolation is utilized to ensure the motion performance of the robot and improve the smooth stability of the robot trajectory. The experimental results indicate that the combination of the semantic laser SLAM system with deep learning and the trajectory interpolation algorithm proposed in this paper yields better graph-building and loop-closure effects in large scenes at SIASUN large scene campus. The use of T-trajectory interpolation ensures vibration-free and stable transitions between target points.

Research on Motion Control and Wafer-Centering Algorithm of Wafer-Handling Robot in Semiconductor Manufacturing.

This paper studies the AWC (Active Wafer Centering) algorithm for the movement control and wafer calibration of the handling robot in semiconductor manufacturing to prevent wafer surface contact and contamination during the transfer process. The mechanical and software architecture of the wafer-handling robot is analyzed first, which is followed by a description of the experimental platform for semiconductor manufacturing methods. Secondly, the article utilizes the geometric method to analyze the kinematics of the semiconductor robot, and it decouples the motion control of the robot body from the polar coordinates and joint space. The wafer center position is calibrated using the generalized least-square inverse method for AWC correction. The AWC algorithm is divided into calibration, deviation correction, and retraction detection. These are determined by analyzing the robot's wafer calibration process. In conclusion, the semiconductor robot's motion control and AWC algorithm are verified through experiments for correctness, feasibility, and effectiveness. After the wafer correction, the precision of AWC is <± 0.15 mm, which meets the requirements for transferring robot wafers.

Research on Small Sample Multi-Target Grasping Technology Based on Transfer Learning.

This article proposes a CBAM-ASPP-SqueezeNet model based on the attention mechanism and atrous spatial pyramid pooling (CBAM-ASPP) to solve the problem of robot multi-target grasping detection. Firstly, the paper establishes and expends a multi-target grasping dataset, as well as introduces and uses transfer learning to conduct network pre-training on the single-target dataset and slightly modify the model parameters using the multi-target dataset. Secondly, the SqueezeNet model is optimized and improved using the attention mechanism and atrous spatial pyramid pooling module. The paper introduces the attention mechanism network to weight the transmitted feature map in the channel and spatial dimensions. It uses a variety of parallel operations of atrous convolution with different atrous rates to increase the size of the receptive field and preserve features from different ranges. Finally, the CBAM-ASPP-SqueezeNet algorithm is verified using the self-constructed, multi-target capture dataset. When the paper introduces transfer learning, the various indicators converge after training 20 epochs. In the physical grabbing experiment conducted by Kinova and SIASUN Arm, a network grabbing success rate of 93% was achieved.

Object Recognition and Grasping for Collaborative Robots Based on Vision.

This study introduces a parallel YOLO-GG deep learning network for collaborative robot target recognition and grasping to enhance the efficiency and precision of visual classification and grasping for collaborative robots. First, the paper outlines the target classification and detection task, the grasping system of the robotic arm, and the dataset preprocessing method. The real-time recognition and grasping network can identify a diverse spectrum of unidentified objects and determine the target type and appropriate capture box. Secondly, we propose a parallel YOLO-GG deep vision network based on YOLO and GG-CNN. Thirdly, the YOLOv3 network, pre-trained with the COCO dataset, identifies the object category and position, while the GG-CNN network, trained using the Cornell Grasping dataset, predicts the grasping pose and scale. This study presents the processes for generating a target's grasping frame and recognition type using GG-CNN and YOLO networks, respectively. This completes the investigation of parallel networks for target recognition and grasping in collaborative robots. Finally, the experimental results are evaluated on the self-constructed NEU-COCO dataset for target recognition and positional grasping. The speed of detection has improved by 14.1%, with an accuracy of 94%. This accuracy is 4.0% greater than that of YOLOv3. Experimental proof was obtained through a robot grasping actual objects.