Location:Home > Students > Past students
Feng Guo

Biography

Enrollment Date: 2013

Graduation Date:2016

Degree:M.S.

Defense Date:2016.05.31

Advisors:Songping Mai

Department:Institute of Microelectronics,Tsinghua University

Title of Dissertation/Thesis:Design of Vision-based Three-dimensional tactile sensor

Abstract:
With the development of technology, application of robots is becoming increasingly widely used. Various fields including precision manufacturing, medical service and scientific experiment call for robot’s ability for dexterous manipulation, while this kind of ability has a high demand for tactile sensing. As a result, the design of tactile sensor with high speed, high precision and high resolution has become a research hotspot. Commercial tactile sensors usually measure tactile information in a capacitive, resistive or piezoelectric way. These kinds of principle usually can only measure normal force, while vision-based tactile sensor can measure not only normal force but also shear force, deformation of contact surface and even texture of contact object. In addition, it also has the advantages of good flexibility, high resolution, no need for many electrical wires and being immune to electromagnetic noise. All of the above draw attention of many researchers. This thesis starts from the principle of visual tactile sensing. We researched on the principle to measure three dimensional contact force and three dimensional deformation of contact surface from markers which are on the surface of an elastic medium. We also measured the normal position of markers by the introduction of DFD method, which was not solved in previous vision-based tactile sensors. Based on the visual tactile sensing principle, we designed mechanical structure of the proposed tactile sensor. As the elastic medium has a great impact to the performance of tactile sensor, we researched on the preparation technique of the elastic medium and the markers. We use PDMS to prepare elastic medium which is transparent and as flexible as human finger. We also use two methods to prepare markers: silk printing and metal sputtering. The process of silk printing tends to be simpler and more mature while metal sputtering can obtain a higher space resolution. We designed the tactile sensing algorithm of the proposed sensor , which includes calibration, image preprocessing, calculation of normal position and shear position of each marker and output of displacement vector and three dimensional deformation. We chose a compact USB camera to capture the markers and completed validation of the algorithm on PC. The accuracy of measurement of shear position and normal position is 0.0015mm and 0.06mm respectively. It was also demonstrated that the proposed sensor can measure the magnitude and direction of shear force and torque from the shear displacement vector of the markers and can reconstruct the three dimensional deformation. The algorithm was realized on PC with a speed of 24fps. In order to make the sensor work independent of PC and improve the working speed, we designed the tactile sensing SOPC based in FPGA. Two set of structure were designed to measure the shear displacement of markers. The first one collects image data by DMA based in a Nios II CPU. It has also got VGA and UART which can improve the efficiency of validation and debug. The other one realized the calculation of shear position in circuit way which in turn makes the speed of the sensor better.