Online Self-Calibration of Cable-Driven Parallel Robots Using Covariance-Based Data Quality Assessment Metrics

Abstract

Abstract This article presents an algorithm to perform self-calibration of cable-driven parallel robots (CDPRs), where the CDPR’s end-effector pose is estimated in conjunction with the calibration of biases in CDPR’s measurements. Two new metrics, known as the position dilution of precision (PDOP) and orientation dilution of precision (ODOP), are introduced as a means to quantify the quality of data collected with regards to self-calibration. These metrics are based on a covariance matrix that is computed online as part of the proposed self-calibration algorithm, which results in the PDOP and ODOP directly corresponding to the standard deviation of the position and orientation errors, respectively. These metrics are used to intuitively select which data points contribute to improved calibration, resulting in a computationally efficient algorithm requiring few data points to maintain accurate calibration. In addition, the PDOP and ODOP provide a means to assess when sufficient calibration data have been collected. Numerical results involving an inverse kinematic simulation with rigid cables and a dynamic simulation with flexible cables indicate that the proposed algorithm is capable of performing self-calibration in a computationally efficient manner. Moreover, the simulation results indicate that the proposed PDOP and ODOP metrics result in smaller position and orientation errors when used to prune the dataset compared to the observability indices found in the literature. Accuracy of the proposed algorithm is also confirmed through experiments when compared to ground-truth pose data.