Robust parallel robot calibration with partial information

Abstract

A new algorithm for calibrating Gough platforms is proposed. It requires internal sensor measurements and only the position information is provided by external sensors. It removes the need to measure orientation, which is intricate and error-prone, by algebraic elimination. This approach, relying on resultant and dialytic elimination, produces an equivalent, yet simpler, set of equations. A numerical simulation is given to compare the existing techniques with our method using partial information, which proves to be significantly more robust, without compromising accuracy. It reduces initial error in pose determination by 99% and 80-98%, in two sets of experiments with realistic conditions. We compare different choices for the measured configurations and show the relevance of configurations at the workspace's boundary. This increases reliability by avoiding to use any random measured configurations.