Abstract: For strapdown inertial navigation system (SINS) in autonomous underwater vehicle (AUV), launch shock or insufficient alignment time may result in considerable initial misalignment errors, and the azimuth misalignment error is a major factor to reduce AUV position accuracy. In this paper, dead-reckoning (DR) and strapdown gyro-compass alignment (GA) algorithms are synchronously executed soon after AUV launch using raw data from strapdown inertial measurement unit (SIMU) and Doppler velocity log (DVL). After a short-time of GA level tuning stage, DR level attitudes are corrected to diminish level errors, but DR azimuth is corrected only when GA is convergent during the azimuth adjustment stage. Azimuth error between GA and DR is calculated, and DR position errors can be estimated by the similarity theory of DR course and AUV actual course. Hence, AUV navigation accuracy can be improved. The simulation results show that the proposed method can perfectly correct position errors due to initial misalignment errors, but excluding position errors caused by DVL scale error. In the simulation case, AUV position errors diminish to 19.8% after error compensation.