Evaluation of Robot Kinematic Calibration Methods with Different Pose Error Expressions and Length Units

This paper investigates the impact of different pose error expressions and length units on the performance of robot kinematic calibration based on the product-of-exponential (POE) formula. Two models are derived: one expressing pose error in the base frame and another in the tool frame. A theoretical explanation is provided to show why the tool frame expression decouples orientation measurement and position errors, making it more robust to metric scaling. Simulations under various noise conditions and experiments with an Aubo i5 robot validate the hypothesis. The results demonstrate that the calibration model using tool frame pose error expression consistently achieves better accuracy and is less sensitive to changes in length units. This work provides significant insights for improving calibration effectiveness by selecting suitable pose error expression and length unit.