Investigations Into Precision Design And Kinematic Calibration Of High-speed Pick-and-place Parallel Robots

This dissertation deals with the key issues relevant to error modeling,tolerance design,parameter identification and error compensation of a class of Delta-like high-speed parallel robots having R-?SS?² limbs with a goal to establish a system by which geometric accuracy of the robotic mechanisms can be ensured.The following contributions have been made.With aid of the first-order perturbation technique,a general and systematic approach is proposed for geometrical error modeling of high-speed parallel robots having R-?SS?² limbs by taking into account encoder offsets of actuators and all possible geometrical source errors.The established error models allow the source errors affecting the compensatable and uncompensatable pose accuracy to be separated in an explicit manner,enabling the processes of tolerance design and kinematic calibration to be integrated into a unified mathematic framework.Building upon statistical analysis,probability density/distribution of composite error of the end-effector is developed.It shows that the mean of the composite error is a nonlinear function of,whereas its second moment is a linear function of deviations of the source errors,leading to the definition of a set of local and global sensitivities useful for investigating the influence of geometrical source errors on the pose accuracy.By minimizing manufacturing cost subject to the specified error confidence level of the end-effector and manufacturing feasibility of the related components,two effective approaches are proposed for optimal tolerance design of the source errors affecting the uncompensatable pose accuracy.The proposed methods for tolerance design have been employed to two prototypes developed by Tianjin University.The experimental results show that the tilt angular errors can be effectively suppressed below1?10^-3 rad via tolerance design,satisfying the prescribed pose accuracy.In this way,a simplified error model can be used for kinematic calibration.A 1-dimensional or distance based error model is formulated for source error identification and pose error compensation.Then,by means of residual proportional index and the principal component analysis?PCA?,the algorithms for optimal measurement configuration selection and robust source error identification are investigated,two important issues for improving the measurement efficiency as well as identification accuracy.A hierarchical strategy is then proposed via rough calibration by the identification of encoder offsets of actuators as well as the fine calibration by that of all possible source errors.Equipped with the proposed error model and the estimated source errors,linear compensators for both rough and fine calibrations are developed for the real time implementation,which have been already built in the CNC controller.The experimental results show that a volumetric/rotational accuracy of 1 mm/±3 deg of the end-effector can be achieved after rough calibration,and can be improved further to 0.4 mm/±1 deg after fine calibration.The outcomes of this dissertation are useful for the establishment of a system to ensure the geometrical accuracy of high-speed pick-and-place parallel robots in the stages of design,manufacturing and application.