Genetic algorithm based parameter identification for parallel manipulators

Kinematic calibration is a process in which the actual values of geometric parameters are estimated so as to minimize the error in absolute positioning. This paper presents a genetic algorithm based method of parameter identification for parallel manipulators. It removes the need to measure orientation, which is intricate and error-prone. The method using the inverse kinematics model, instead of the forward one which faces more serious stability issues since it relies on a numerical solution of forward kinematics with a problematic convergence in the presence of noise, is more robust because it decouples the problem into the calibration of each leg independently, and the constraints admit an analytical form.