Objectives  In order to improve the combined navigation accuracy of USVs by strapdown inertial navigation (SINS) and doppler range finder (DVL), the DVL error needs to be accurately calibrated.

  Methods  this paper carries out error modeling of the installation error angle, lever arm, and scale factor of the DVL, and focuses on the analysis of the impact of the lever arm error term on the combined navigation positioning in the vessels, pointing out the necessity of its calibration; proposes a DVL error calibration method based on the parameter estimation between two point sets, which treats the SINS/GNSS navigation parameters and DVL output parameters as two point sets, converting the calibration problem into an estimation problem of the difference between the two point sets, and using the Kalman filter to estimate the error parameters between the two point sets; using the SVD（singular value decomposition）-based observability analysis to quantify the observability of the filter under different motion conditions, giving the motion recommendations of the carrier during the calibration process.

  Results  The results of mathematical simulation and real ship experiments show that the global relative accuracy of the calibrated SINS/DVL algorithm can reach a voyage of 0.122%; the positioning error is smoother after compensating for the lever arm, and has a smaller positioning error overall.

  Conclusions  In the case of angular maneuver, the lever arm causes abrupt changes in SINS/DVL positioning, and the compensation of the lever arm can make the positioning error smoother; the proposed method provides a feasible way to calibrate each DVL error including the lever arm.