Segmented Underwater Terrain Matching Based on Q-Learning

Underwater terrain matching is one of the important technologies for autonomous underwater vehicles (AUV) to achieve long-distance autonomous and precise positioning, and it is also one of the key technologies for ocean exploration. Because the terrain matching method similar to TERCOM (S-TERCOM) only considers the translation in underwater terrain matching process, in this paper we propose a segmented underwater terrain matching method based on Q-learning (S-UTMQ) to address this issue. The main idea of the proposed S-UTMQ method is to comprehensively consider the elevation data before and after the positioning nodes, and dynamically adjust the angle range of each positioning node for matching in real time based on Q-learning, so as to make the terrain matching process more accurate and efficient. The simulation results show that the proposed S-UTMQ method is more effective than the S-TERCOM in the case of different segment positioning lengths, different course angles and curves of trajectory, and has more advantages in accuracy and efficiency.