A local POE-based self-calibration method using position and distance constraints for collaborative robots

Abstract

Collaborative robots are designed to not only work alongside humans but also adapt to new tasks quickly. To improve their absolute accuracy, self-calibration methods utilizing portable measurement devices are cost-effective solutions. Based on the local frame representation of the Product-of-Exponential (POE) formula, a new modular kinematic error model is derived for collaborative robots, in which all geometric errors in a joint-link assembly, i.e., a dyad, are lumped into a local rigid body motion applied to its nominal joint axis. Its major advantage over the conventional local POE-based kinematic error model is that all columns in the error transformation Jacobian matrix are functions of joint displacements, which is critical to formulate singularity-free self-calibration models based on geometric constraints. A portable and cost-effective self-calibration device with both position and distance constraints is proposed for collaborative robots, which can readily collect a wide range of constraint errors by drag teaching within a collaborative robot’s workspace. Additionally, a two-step self-calibration method is established for both the robot arm and its base calibration. Its effectiveness and robustness are validated through simulations and experiments. The experimental results on Aubo i5 show that its average position error is significantly reduced from 4.11mm to 0.36mm after calibration.