Research On Humanoid Robot Servo Grasping Objects Based On Kinect

With the arrival of science and technology revolution,robot technology has become one of the most core technologies in the competition of overall national strength among countries in the world.Robot Grasping Technology,as one of the most important technologies in the field of Robotics,is the most basic technology.Different from industrial robots,the automatic grasping ability of humanoid robots is mainly affected by the environment and synergistic sensor.What' more,the diversity and randomness environment also put forward higher requirements for the sensory system of humanoid robot.In fact,in the real experiments on NAO,the monocular vision is diffict to gather the depth information of the environment;and the overlapping field of view of the binocular vision is small,which limits the grasping scope of NAO.Therefore,the paper presented a solution that robots perceive the outside world through the Kinect senor instead of the camera,and then do a research on the kinematics of the robot,the identification and localization of the target,motion path planning of the manipulator and the real-time object-grasping operation.The main contents of this article are as follows.(1)According to the robot kinematics,the relevant theoretical knowledge were introduced,and the mathematical model of NAO's arms based on D-H and kinematics forward equations were established,then the inverse kinematics equation were solved by using analytical method,which provides a foundation for the realization of NAO robot grasping target.(2)For the realization of high-precision target localization of the robot,This paper presented a design scheme of humanoid robot target localization based on Kinect.Firstly,carrying out the image processing of the scene information from Kinect to get the coordinates of the target center point;then the coordinate systems for the Kinect and the robot were established,the Bursa coordinate transformation model between Kinect and the robot was constructed,and the unknown parameters with the model were solved by the linear total least squares algorithm(LTLS)algorithm.After that,to realize the robot target localization,transforming the target center coordinates of Kinect to the robot coordinate system through the model.At last,the experiments were carried out on the humanoid robot NAO,which shown that the proposed target localization is reasonable and feasible;simultaneously,NAO robot can locate the object in real time,reliably with the method,which satisfied the localization accuracy required with the robot to grab the object.It is also more accurate than the monocular vision of NAO robot.(3)In order to make robots quickly and accurately grasp the object,after analyzing the working space of the robot,the iteration method and the geometric method were used to calculate the motion reachable space of NAO's arms Aiming at the characteristics of multijoints and high-dimensional space of the NAO robot,this paper planed the motion path of the robot arm via the improved Rapidly-exploring Random Tree(RRT* algorithm),which introduced the grid search method and the bidirectional expansion strategy into the original RRT* algorithm.The experimental results show that this method not only preserves the characteristics of RRT*,but also improves the efficiency of the path planning.(4)Under the constraints with obstacles and without obstacle,through the construction of the overall robot self-crawling system,by combining the above(1),(2),and(3)expounded methods,then achieving the goal of the robot autonomy graspes diverse objects in different environments.At the same time,the experiments verify the correctness of the theory used in this paper.