Research On Mobile Manipulator Grasping Object Based On Binocular Vision

As the base of the traditional robot is fixed on a certain position, the motion space of robot is limited, however, the mobile manipulator consists of the mobile platform and the manipulator with multi-degree of freedom, the mobile platform makes the mobile manipulator move vastly and at the same time the manipulator with multi-degree makes the mobile manipulator executes flexibly. The degree of the mobile manipulator contains the degree of both mobile platform and manipulator, and there are strong disturbances and uncertainties, so the control of the mobile manipulator's motion planning is difficult and the mobile manipulator is the hot spot and focus of current research with broad application prospect.This paper adopts a motion planning method, which considers the mobile platform and the manipulator with multi-degree separately to control the mobile platform and the manipulator with multi-degree respectively via decision-making layer in the motion planning progress. The mobile platform which this paper adopts is the two-wheeled mobile robot, which is a synthetical system including a variety functions of environmental perception, dynamic decision-making and planning, motion control and perform, etc. The motion control problems about two-wheeled mobile robot mainly includes point stabilization control, path planning and trajectory tracking. For manipulator with multi-degree, we could utilize feature deviation directly to construct closed loop controller in the image space based on uncalibrated visual servo.This paper mainly does research to realize the task, which requires the mobile manipulator to move to catch the target object automatically, the main contents as follows:(1) design and complete the motion control to let the mobile manipulator catch the target object based on milemeter and uncalibrated visual servo. The method includes three steps, the first step is that, the two-wheeled mobile robot moves near the target object via milemeter, design the milemeter motion control system based on the feedback of the position and posture of the two-wheeled mobile robot, realize the location and motion control of the two-wheeled mobile robot.The second step is that, adjust the position of the two-wheeled mobile robot based on the information from the binocular vision system with special two-axis parallel layout, to make sure that the two-wheeled mobile robot faces the target object in aforward direction with a certain distance.The third step is that, choose the point with approximate decoupling of location as the image feature and construct the image feature mapping model with qualitative and quantitative description based on uncalibrated visual servo control, use the D-H parameters method to establish the model of the manipulator, combines the analytic method and genetic algorithm to solve inverse kinematics of manipulator, design multi-model visual servo controller to realize the transformation between the Cartesian space and the joint angle space, so the manipulator could catch the target object, and prove the convergence of features points.(2) Construct the two-wheeled mobile robot simulation platform and the manipulator with the binocular vision system with special two-axis parallel layout simulation platform, realize the two-wheeled mobile robot point stabilization simulation experiment and the manipulator to catch the target object simulation experiment in Cartesian space.(3) Combines the two-wheeled mobile robot system based on the Maxon brushless dc motor system and DSP controller, the binocular vision system with specialtwo axis parallel layout system and Yixuetong six-degree manipulator system, to construct mobile manipulator system with binocular vision, and realize the experiment the mobile manipulator to catch the target object in cartesian space.This paper proves that via simulation and physical experiment,the location of the two-wheeled mobile robot based on milemeter and the uncalibrated visual servo based on binocular vision system with two axis parallel special layout, could well completes to catch the target object in cartesian space automatically by vision, without solving the image Jacobin matrix.