Font Size: a A A

Spherical Robot With A Manipulator

Posted on:2010-04-18Degree:DoctorType:Dissertation
Country:ChinaCandidate:W ZhuangFull Text:PDF
GTID:1118360308962201Subject:Mechanical and electrical engineering
Abstract/Summary:PDF Full Text Request
The spherical robot with arms is different from the present spherical robot. It is a novel structure of robot, which combines the advantages of general spherical robot with traditional manipulator robot. This type of robot has excellent performance, including high maneuverability to the environment and omni-direction movement, so it will have far-ranging application perspective in various fields. In this dissertation, all research work is mainly on a newtype of spherical robot with arms, but for the purpose of convenient, the platforms of 1 prismatic joint and 5 revolute joints (1P5R) flexible manipulator and 3 revolute joints (3R) flexible manipulator are under consideration as well. Some critical technique about this spherical robot is discussed deeply and detailly in this dissertation. The innovational work includes deriving the dynamic model of the spherical robot, presenting an approach to identify inertial parameters of the underactuated spherical robot, designing an intelligent controller for flexible joint manipulator, and proposing an approach to identify joint parameters of the flexible joint robot under movement. The detail specification of this work will be shown as follow:At the beginning of the dissertation, a novel structure of spherical robot with arms was introduced. This robot has 13 rigid bodies and 15 degrees of freedom (DOF). Kinemetic characteristic of the robot was detailly analyzed. Regarding the sphere shell of the robot as the base and the driving installations inside the shell plus the sphere caps and the arms inside the caps as the links, law of motion of each link was analyzed. Considering the condition of nonholonomic constraints fully, partial velocity, partial angle velocity, generalized active force and generalized inertia force of each link were derived respectively. Based on Kane's equation, dynamic model of the spherical robot was derived ultimately. Joints forces/torques and Euler angles of the robot were analyzed respectively after rectilinear motion and sigmoid curve motion simulation.Secondly, according to the characteristic of the dynamic model of spherical robot, an approach based on inertial measurement unit (IMU) to identify inertial parameters of the underactuated spherical robot was presented. Considering the condition of asymmetrical mass distribution in each link of the spherical robot, dynamical model was constructed with Kane's method, and accordingly, the inertial parameters identification model was derived. Aming at a newtype of spherical robot, the proposal identification method was examplified by an identification experiment according to the information from IMU fixed on the frame of the robot. The experimental result shows that the identification datum are reasonable, and the proposal approach is simple and feasibly, which is a good guidline for inertial parameters identification of other underactuated robots.Thirdly, based on the assumption that the arms system of the spherical robot can be taken as the manipulator system of prismatic joints and revolute joints when the sphere shell was immobile, the dynamic model of rigid-flexible coupling system of a 6DOF flexible joint manipulator was established.According to the dynamic model, a feedback linearization controller of 1P5R manipulator was discussed detailedly. Furthermore, the trajectory was designed in workspace of the system, and tracking controls under the condition of two different motions at the same route were simulated numerically with MATLAB. This research is a good guidance for rationally planning the trajectory of manipulator endpoint.Fourthly, aiming at the multiple-links flexible joint arms of the spherical robot, a sliding mode controller for manipulator was proposed based on Gaussian Radial Basis Function Neural Networks (GRBFNN). Regarding the switching function of each joint as the input and the controlling variables as ouput of RBFNN, the proposed controller was realized by continuous RBF neural network using the approximation ability of the network. The simulations of tracking control of three-links flexible joint manipulator was done with the proposed controller, and the linear quadratic regulator (LQR) tracker and traditional sliding mode controller (SMC) comparely.Finally, focusing on the time-variant joint surface parameters of the flexible joint manipulator, a new approach to identify joint parameters of the flexible joint manipulator under the condition of moving states was presented. Regarding a flexible joint as an elastic twist axis, and combining the wave theory used in structures analysis with rotation transform matrix of joints, the wave equations of mechanism were established. Considering the condition of generalized forces balance and displacement boundary of each joint, the stimulator prediction model of the system under the condition of moving states was derived. Vibration equations were derived according to the predicted stimulator and the measuring twist angles of joints. Joint dynamic stiffness and damping were identified via solving the former identification model by use of least square method (LSM) or neural network method. Experiments on a 3-DOF manipulator were performed when driving the manipulator under two different trajectories.
Keywords/Search Tags:spherical robot, dynamic model, inertial parameter, flexible joint, joint stiffness, identification, sliding mode control, radial basis function neural network
PDF Full Text Request
Related items