Research On Application Of Desktop Robotic Arm In Chemical Experiment

Robotic arms are now being widely used in industrial automation production,bringing more convenience to industrial intelligence.However,different from the working scene of the traditional industrial robot arm,there are many disturbances in the chemical experiment scene,such as unstructured,dynamic,and inaccurate placement of objects.Compared with the industrial robot arm,the desktop robot arm has the advantages of more flexibility,compactness and low cost.The desktop robotic arm can be re-developed for different usage scenarios,which can make it more intelligent.The main problems of how to apply the desktop robotic arms in the chemical experiment environment are whether the positioning of chemical experimental vessel can be identified by depth camera,how to identify and classify the solution in the chemical experimental vessel,and whether the trajectory planning after the robot grabs the experimental instrument can ensure the smooth and accurate movement of the robotic arms.In order to solve the problem of autonomous recognition,grasping and operation of the desktop robot arm in the chemical experiment scene,this paper uses the desktop robot arm as the platform,combined with software and hardware technology,deep learning algorithms,and robot arm trajectory planning algorithm to build the machinery in the chemical experiment scene.A system for automatic arm grasping operation.The main research content of this paper includes the following parts:(1)Aiming at the identification and positioning of chemical experiment containers and other items,this paper proposes a scheme of indirect identification and positioning of chemical containers by means of target detection algorithms to identify chemical labels on chemical containers.This paper first compares different target detection algorithms for algorithm selection;then,for YOLOv3 tiny algorithm,the target detection algorithm is improved for defects with insufficient accuracy of small target recognition,and shallow convolutional layers and residual units are added,and the activation function is changed to Mish;Then,the target detection algorithm is trained and evaluated on the self-built chemical label sticker image data set.The improved algorithm has improved recognition accuracy while ensuring lightweight,and has realized the identification and positioning of chemical experiment containers.(2)Aiming at the problem of chemical experiment reagent classification,this paper uses the method of recognizing handwritten words on chemical label stickers to classify chemical reagents.First,through the research on text detection and handwritten Chinese character recognition algorithms,the method of text detection and recognition using CTPN+CRNN+CTC algorithm is determined.In order to allow the model to run on embedded devices,the text recognition algorithm was lightened and improved,and the convolutional layer was replaced with Mobile Net V3.Secondly,this paper selects single-character images containing common chemical names from the HWDB handwritten Chinese character data set images,and performs image stitching to establish a handwritten Chinese character image data set about chemical reagent names.Finally,the training evaluation of the Chinese character recognition algorithm for handwritten chemical reagents is carried out.The experimental results show that the improved algorithm guarantees the accuracy rate and the lightweight of the model,and realizes the classification of chemical experimental reagents.(3)The trajectory planning algorithm of the robotic arm is studied for the movement speed and acceleration changes of the robotic arm in the action of pipetting and dumping in the experimental operation of the robotic arm grasping the test tube.The comparative experimental test of the trajectory planning algorithm cubic polynomial interpolation and the fifth-degree polynomial interpolation algorithm,the test results show that the fifth-degree polynomial interpolation algorithm can ensure that the acceleration and speed parameters of the desktop robot arm move more smoothly,which can meet the requirements of the robot arm for chemical experiments.The requirement for smooth movement.(4)Build a vision-based autonomous grasping motion system for robotic arms,and conduct system tests around chemical experiment operations.First complete the internal and external parameter calibration to obtain the depth camera parameter matrix,use the ROS system to complete the communication settings and model deployment of the various modules between the systems,and build the robotic arm visual grasping system;then the autonomous grasping operation of the simulated environment is carried out for the chemical experiment.The experimental results show that it is feasible to use a desktop robotic arm for chemical experiments.