Cooperation Detection and Tracking of Underwater Target via Aerial–Surface–Underwater Vehicles

Detection and tracking of underwater target play an important role in enhancing the marine sensing ability. Currently, the above mission is usually conducted in a single platform, and there is a lack of necessary cooperation between different platforms. This article employs unmanned aerial/surface/underwater vehicles (UAV–USV–UUV) to develop a cooperation detection and tracking solution for underwater target. We first use the measure theory to construct a heterogeneous detection mode, such that the target detection probability can be maximized by adjusting the formation shape between USV and UUV. After the target is detected by the UAV–USV–UUV networks, a deep learning algorithm called <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">d</u>epth <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">d</u>eterministic <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</u>olicy <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">g</u>radient (DDPG) is designed for UUV to track the trajectory of target. In order to guarantee the communication connectivity among UAV, USV, and UUV, a multistep location prediction strategy is incorporated into the tracking procedure. Note that the advantages of our solution are highlighted as: 1) the cooperation of UAV, USV, and UUV in this article can improve the detection probability over the single platform system; and 2) the DDPG-based tracking algorithm in this article is more efficient for handling continuous complex underwater environment as compared with the deep Q-network. Finally, simulation and experimental results are both presented to verify the effectiveness of our solution. As such, our solution is more useful for the marine engineer to remotely sense the ocean from the communication and control viewpoints.