Differential-Evolution-Based Parameter Identification Method for a Redundant Hybrid Robot Using POE Model

摘要

This paper presents a kinematic calibration method for a redundant serial-parallel robot to improve its end-effector positioning accuracy. The studied robot is composed of a kinematically redundant serial mechanism with four degree-of-freedom (DOF) and a hexapod parallel manipulator with full six degree-of-freedom in a three-dimensional space. The serial mechanism is designed to enlarge the robot's workspace whereas the parallel manipulator is used to improve the robot end-effector accuracy. The robot will be adopted for the assembly of vacuum vessel of the international thermonuclear experimental reactor (ITER) by carrying out various tasks, such as welding, machining, non-destructive testing, measuring the gap between two adjacent sectors and transporting a premade splice plate to match the measured gap. Based on the product of exponentials (POE) formula, an error model involving 60 kinematic parameters is derived, which accounts for the kinematic errors originated from the manufacturing and assembly processes. Due to its hybrid serial-parallel kinematic structures and a large number of identification parameters in a high nonlinear error model, the traditional iterative least-square algorithms cannot be used to identify the error parameters. In this paper, we proposed a hybrid calibration method for serial-parallel hybrid robot by combining the forward and inverse kinematics together. The parameter identification process is transformed into a global nonlinear optimization problem, and then Differential Evolution (DE) algorithm is employed to search a set of optimum solution in the error model to minimize the derived objective function. Numerical simulations reveals that all the preset error parameters can be successfully recovered under the ideal conditions where the measurement system is assumed to be perfect without measurement noise. Simulations for the imperfect conditions with measurement noise also demonstrate that the identification method is robust and effective.