This paper proposes a mission management system (MMS) to manage target recognition and path planning for the autonomous underwater vehicle (AUV) ocean survey. The system is autonomous and can trigger the corresponding mechanism according to the real-time ocean environment. The purpose is to survey only interest areas after online targets recognition, reducing wasted time optimizing mission-independent paths. The recognition part uses an improved end-to-end lightweight neural network algorithm, wide perception based on ShuffleNet (W-ShuffleNet), which enhances the image of the carried sensor information and performs online recognition, the accuracy of the simulation test is above 98%. The path planning part uses a real-time path planning (RTPP) method to explore the best strategy by evaluating the recognition results over a while to achieve the purpose of real-time adjustment of the AUV path, which effectively saves time and energy costs. This system is essential for improving overall mission performance and conducting effective ocean surveys under realistic conditions. We conducted sea trials with an AUV equipped with side-scan sonar, and the sea trial results proved the effectiveness of our proposed MMS.
展开▼
