Motion difference analysis between robot arm and human arm by human–robot metrology system based on inertial measurement unit sensors

The development of robotic arms for high-precision applications requires a detailed understanding of motion discrepancies between human and robotic limbs. This paper proposes a human–robot metrology system for analyzing and differentiating dynamic motion characteristics using inertial measurement unit data. The proposed method employs batch-based temporal segmentation of time-series signals combined with statistical descriptors and dynamic time warping (DTW) to localize fine-grained motion differences, in contrast to conventional global analysis approaches. Experimental results show that the proposed framework effectively captures localized variations in acceleration and angular velocity, revealing higher vibration levels and irregular motion patterns in the robotic arm compared to human movement. Compared to conventional point-to-point metrics such as Euclidean distance, the DTW-based approach provides a more accurate assessment of motion similarity by accounting for temporal misalignment. Statistical analysis confirms significant differences between the two methods (p < 0.001), with large effect sizes (|d| > 1.7), indicating substantial overestimation of motion discrepancy by Euclidean distance. Frequency-domain analysis further identifies increased noise components in robotic motion, while magnetic field distortions caused by structural materials significantly degrade yaw estimation accuracy. This research demonstrates that segmental metrology provides more sensitive and informative motion characterization, highlighting its potential for improving robotic arm control and calibration, particularly in environments affected by magnetic interference.