Accuracy of position and orientation for consumer-grade tracking sensors with Hand-Eye Calibration

Abstract

Extrinsic calibration of motion tracking sensors is a fundamental aspect for the design and development of wearable measurement system and robotics devices. In this study, we evaluated the accuracy of consumer-grade tracking sensors, implementing a linear hand-eye algorithm, in order to minimize the systematic effect of sensor-to-frame misalignment. Three consumer-grade tracking sensors were used in this study, an Inertial Odometry-Vision system, a VR-device based on infrared (IR) structured light, and an Inertial Measurement Unit. The three systems were placed on the end-effector of an industrial robot UR5e, whose trajectories and rotations were taken as reference. The hand-eye-calibration method was applied to estimate the actual alignment of the sensor’ frames in respect to the end-effector one. The experimental protocol was conducted considering a set of 16 calibration trials, with two joint speeds: (i) 30°/s (acceleration 30°/s) and (ii) 15° (acceleration 15°/s). Each run had a time duration of 3 min. The 8 low-speed trials were characterized by a reproducibility of ±1°, ±1.2° and ±1.8°, for the three sensors, respectively. The beneficial effect of calibration resulted significant only for the Inertial Measurement Unit, with an inaccuracy reduction of 2°. No significant improvement in the accuracy of the other sensors was observed for either orientation or position tracking. The results suggest that for applications where a precise alignment among sensors is required, the accuracy of consumer-grade tracking sensors may be not sufficient, even implementing an optimal calibration procedure, in particular for Inertial Measurement Units.