Geomagnetic Navigation for AUV based on Deep Reinforcement Learning Algorithm

For the problem of geomagnetic navigation in the case of autonomous underwater vehicle (AUV) without a prior geomagnetic information, this paper proposes a geomagnetic navigation algorithm based on deep reinforcement learning. Using the correlation between the geomagnetic parameters and the navigation trajectory, the Deep Q Network is used for choosing the navigation direction by multiple geomagnetic components, and the relative position is estimated by the difference between the geomagnetic parameters of the current position and the target position. Under the influence of the above navigation model, the multiple geomagnetic component gradually approaches the target value with the movement of the aircraft, thereby achieving the navigation purpose. By comparing with the evolutionary algorithm (EA), it is proved that the DQN algorithm have better convergence ability.