Optimization of Sensing Locations of Autonomous Underwater Vehicles for Optimal Environmental Prediction and Acoustic Target Tracking

A distributed, mobile, cooperative, and autonomous sensor network formed by a collection of autonomous underwater vehicles equipped with various types of environmental and acoustic sensors is an important type of tool in many underwater monitoring and surveillance applications. optimizing the spatial-temporal sensing locations of the networked autonomous underwater vehicles to enable the vehicles to collect informative spatial-temporal data streams for accurate predictions of the states of the dynamic underwater environments and maneuvering targets is a key issue for the efficient operation of such robotic sensor networks. This paper presents a developed software tool to implement the optimization. In the developed tool, a deep reinforcement learning-based strategy is developed to optimize the sensing locations of a swarm of autonomous underwater gliders for the accurate prediction of three-dimensional states of ocean environments, where the prediction is implemented with a data-driven ocean model based on the dynamic mode decomposition method with the sparse glider sensing data. And a Monte Carlo tree search-based strategy is developed to optimize the motion path of an autonomous underwater vehicle for detecting and tracking a target with sonar sensor, where the ensemble forecasts of environments are taken into account in the prediction of sonar performance of probability of detection and the path optimization. This paper presents the two developed strategies of optimal sensor placement and path planning and the simulation demonstrations showing their performance.