Position-Based Visual Servoing Of Cable-Driven Parallel Robots

Cable-driven parallel robots(CDPRs)are new type of parallel robot that adopts flexible cable instead of rigid linkage.CDPRs provide large workspace potentially,have high payload-to-weight radio,and can be assembled/reassembled and reconfigured easily.From the control point of view,the workspace control of the CDPRs is superior to the joint space control,but the parallel robot has no closure form direct kinematic solution just like the inverse kinematics problem of the serial robot,which makes the workspace control for CDPRs very difficult.In addition,the cable can only pull,the error in the control process leads the cable loose,so the absolute control accuracy of the CDPRs depends more on the precise robot model,especially the redundantly actuated 8-6 CDPRs.Therefore,this dissertation considers the special characteristics of the CDPRs,and taking the 8-6 CDPRs as the research object,the PBVS(Position-Based Visual Servoing)control and the workspace dynamics PBVS control were studied respectively.The main work is summarized as follows:(1)For the problem that the parallel kinematics has no closure form direct kinematics solution,an external measurement method is used to measure the pose of the platform in the inertial reference frame using a vision sensor.For the significant delay problem of existing pose estimation algorithms,the visual sampling period is much larger than the motion control period and cannot be directly applied to motion control.Consider adding short-term prediction within the visual sampling interval.In the process of visual servoing control,the image processing time and the iterative estimation convergence time of pose are uncertain,leading to the visual sampling period is not constant,while the control period of the workspace planning trajectory is constant.Considering the above two problems at the same time,a multi-threaded real-time measurement scheme for pose estimation and position prediction is designed.(2)Considering the trajectory tracking control problem of the moving platform under low-speed motion,a first-order approximation model is used to model the motion of the moving platform.The short-term prediction algorithm in visual real-time measurement is based on the use of linear Auto-Regressive(AR)prediction algorithm.Considering the low speed motion situation of the moving platform,the visual servo interaction matrix of the 8-6 CDPR was deduced,and the design of the velocity resolved controller was designed.Based on the above measurement scheme and control scheme,the workspace kinematic PBVS control of the 8-6 CDPR was designed and verified on the actual robot platform.(3)In order to control the trajectory tracking of the moving platform under high-speed motion,the coupling characteristics of dynamics must be taken into account to further complete the dynamic visual servoing control.Firstly,a Newton-Euler method is used to establish the workspace dynamics model of the robot.Then,considering the high-speed motion conditions of the moving platform,the motion model of the moving platform is established based on its dynamic model,and the short-term prediction scheme is the Extended Kalman n-step prediction algorithm.Finally,based on the above specific real-time measurement scheme and robot dynamics model,using the typical computed torque controller in the robot control,a complete workspace dynamic PBVS control scheme of 8-6 CDPR was designed.Simulation verification of visual measurement scheme was completed.