Research On Tactile Perception And Object Classification For Soft Grippers

With the wide application of robots,the rigid grippers of traditional robots cannot meet the requirements of grasping irregular and easily damaged objects.The emergence of soft grippers solves the defects of poor adaptability and large rigidity of rigid grippers.With the continuous intellectualization of robots,how to make robots perceive the properties of the objects they grasp and the surrounding environment is an important research direction.Tactile perception can not only directly measure the physical properties of the objects they touch,but also can adjust the robot through feedback.This thesis takes the soft gripper as the object,designs the flexible tactile sensor,and studies the structure design of the tactile sensor array,the design of the tactile information acquisition system,the manufacture and analysis of the soft gripper with different configurations,the manufacture and control of the tactile sensor gripper,the construction of the object classification test platform and the analysis of the experimental results on the premise of ensuring that its performance meets the follow-up research through a number of performance tests.The specific research contents are as follows:Firstly,this thesis analyzes and compares a variety of tactile sensing materials and tactile sensors with different sensing principles,and analyzes their working principles and advantages and disadvantages.Select Velostat material as the sensing element for tactile sensors and explain the sensing principle of this material.Produce and analyze the advantages and disadvantages of beak type and fin type soft claws,and discard beak type soft claws that cannot provide installation surfaces.And explain the relevant theories of classification algorithms.Then,in order to solve the problem of initial tactile pressure in the production method of traditional piezoresistive tactile sensors,a corresponding flexible circuit board(FPC)was designed based on the tactile sensing array structure,integrating the protective film,wire,and signal transmission port.Model the circuit structure of the sensor array and use Arduino Mega2560 to design and manufacture the corresponding signal acquisition system.Design the corresponding amplification circuit using LM358 N,make it into a PCB,and reserve signal transmission ports.To visually display the distribution of tactile information,MATLAB software is used to convert tactile data into grayscale images.Next,use the finned soft gripper to complete the production of the tactile sensing soft gripper,and use the DRV8825 driver chip to build the control system of the finned soft gripper.And through four performance testing tests,including sensing sensitivity,loading stability,repeatability,and consistency,the sensor has a significant lag phenomenon;Continuous and stable response,with data attenuation less than 5% within five minutes;Conduct 140 cycle tests on the sensor,and the sensor response is stable;The response curves of the three sensor samples are basically the same.Finally,use the Lib SVM toolbox as a classification algorithm tool and build a tactile classification experimental platform using various aluminum profiles.Collect a dataset of1422 labels from nine different objects,and use Support Vector Machine(SVM)to classify tactile data from different claw heads.Finally,the classification accuracy of left finger tactile data,right finger tactile data,and double finger tactile data was 81.1%,84.5%,and 91.1%,respectively.