Geomagnetic Navigation with Adaptive Search Space for AUV based on Deep Double-Q-Network

In order to solve the problem that the AUV uses geomagnetic navigation for multiple round trips between the two places without a geomagnetic database, a geomagnetic navigation algorithm with adaptive search space for AUV based on Deep Double-Q-Network is proposed. The algorithm takes multiple geomagnetic parameters as input and uses the DDQN reinforcement learning algorithm to predict the heading angle. In addition, a boundary function is constructed to limit the search space of the AUV, so that the AUV can search the target point faster. Under the effect of the DDQN navigation algorithm, as the number of AUV round trips increases, the number of search steps gradually decreases. Finally, AUV reaches the target position with fewer steps. Simulation experiments prove the effectiveness of the algorithm.