Abstract: Aiming at the problem that there was a lot of noise in the measurement of the direction of the gyroscope, resulting in a large error to the dead-reckoning navigation, we proposed a method of dead\|reckoning navigation based on neural network. The method used neural network to investigate  the time-varying relationship between acceleration measurement and orientation measurement. Thus, we used acceleration value to calculate direction value through this timevarying relationship without using the gyroscope, and then completed the autonomous underwater vehicle (AUV) dead\|reckoning navigation. The results show that the algorithm can  only use the accelerometer to complete dead\|reckoning navigation. So we can avoid dead-reckoningerrors caused by gyroscope noise at the sea surface. Simulation results show that the algorithm has higher accuracy.

Key words: neural network, deadreckoning, accelerometer