Research On Bionic Hand-eye Coordination Control Of Service Robot

With the rapid development and wide use of service robots, the requierments onservice function become higher and higher. In recent years, inspired by neurobiologyand biological vision, biologically motivated hand-eye coordination controltechniques became a hotspot in the field of robotics research, and many scholars, athome and abroad, did lots of researches on it.The research carried out in this dissertation was supported by the key projectsfrom the State High-Tech Development Plan (863program) of China, and the keyprojects from Shanghai Science and Technology Commission. All the research workwas done on the service robot made by my Lab..This dissertation is mainly focused onthe research of target object identification and localization, static target grasping andmoving target tracking. They are described in detail as following.(1) The structure and components of the service robot were introduced in detail inthis dissertation. The service robot is an omni-directional mobile robot, equiped withtrinocular stereo vision and modular arm. The arm is light, flexible, and can becontrolled to grasp static target object, and track moving target object as well, underthe guidence of stereo vision. The service robot is intelligent, open-coded, andreconfigurable. It shows the characteristics of modern service robot.(2) The image edge feature extraction algorithm based on Contourlet HiddenMarkov Tree Model (HMT in short), and the feature matching algorithm based on theimproved normalized cross-correlation were studied in this dissertation.In recent years, although lots of image edge feature extraction algorithms wereresearched, most of them couldn’t give a optimal expression for the high-dimensionalfunctions including line, or surface singularity. However, the function with a line orsurface singularity is very common in the high-dimensional space. For example, thediscontinuity of smooth boundary of some object is often shown as a line singular onthe smooth curve. In order to remedy the deficiency, the method on direcectionalmultiscale image edge feature extraction based on Contourlet HMT, was studieddeeply in this dissertation. This method is based on image multiscale geometricanalysis theory and human visual characteristics. This method provides the optimalrepresentation of high dimentional function, and is effective for the edge detection ofhigh dimensional image, because of its characteristics of directionality, anisotropy, multi-resolution, and time-frequency like wavelet. With the superiority in capturingdirectional information, contourlet HMT gives satisfactory recognition performancefor multi-directional components (such as circular and irregular shapes), and will bewidely used in the field of service robot for target object recognition.About feature matching, although it is accurate and anti-noise to match with thenormalized cross-correlation algorithm, the coefficient is difficult andtime-consuming to be gotten. It is necessary to simplify the calculation of coefficientso as to improve the speed of matching. In order to solve this problem, theconvolution was used to calculate the coefficient. By this way, the calculation speedwas improved highly(3) In order to get the position of the target object, the algorithm of stereomatching was researched in this dissertation. According to the imaging mechanism ofhuman binocular steero vision, the matching algorithm was proposed based onepipolar geometry and homography constraints. The accuracy of localization wasimproved greatly by the algorithm.(4) The optimal trajectory of human hand movement was analyzed, and themathematical model of it was given. Based on the mathematical model, kinematicmodel of the robot arm was established, and the bionic path, along which the robotwas controlled to grasp target object, was planned. The robot can be controlled tograsp target object along straight line trajectory, and its speed curve is bell-shaped.According to experiment, the service robot can be controlled to grasp the target objectaccurately and smoothly along the bionic trajectory.(5) As for the dynamic target object tracking of the service robot, the bionic pathwas planned with the mathematical model of human hand-eye coordination.According to the experiment, the sevice robot can be controlled to move smoothly totrack the target object.The research work done in this dissertation is very helpful to improve thecapability, and expand the application fields of service robots.