A Two-Step Method for Synchronization of Low-cost GNSS and Visual-Inertial Sensor Timestamps

The integration of global navigation satellite system (GNSS) and visual-inertial navigation system (VINS) has received considerable attention recently. GNSS sensors and visual-inertial measurement units exhibit complementary characteristics. By leveraging the strengths of GNSS in global positioning and VINS in providing smooth and continuous trajectories, the combination of GNSS and VINS can mutually enhance each other. To effectively fuse data from these different sensors, strict synchronization of the sampling times is crucial. However, in low-cost GNSS and VINS sensors, the measured timestamps are sourced from separate systems, resulting in unstable time delays (time offset) between the timestamps. In the absence of hardware synchronization, these delays become completely unknown, leading to unpredictable impacts on positioning accuracy. Therefore, this study proposed a two-step method to calibrate the time offset between GNSS and VINS sensor. In our approach, the temporal offset was calibrated by estimating the delay in a cascading manner. First, time-differenced carrier phase (TDCP) and VINS velocity data were loosely combined, effectively constraining the time delay within the sampling period of the camera. Subsequently, an optimization process in the second step further reduced the error in the time delay. Experimental results demonstrated that the extracted time offsets achieve an accuracy of 1 ms. With our method, comparable positioning accuracy was achieved without hardware synchronization. Furthermore, our approach provided a universal and cost-effective solution for the online calibration of time delays between GNSS and high-precision odometry.