Solving the AUV Navigation Problem on a Real-Time Operating System

The paper is devoted to the problem of determining the coordinates and orientation angles of an autonomous underwater vehicle (AUV) relative to a stationary landing platform using a high-frequency short-range hydroacoustic system. The AUV is equipped with receiving antennas, which are located in the stern and bow of the vehicle, as well as four emitters installed along the edges of the platform and emitting a non-directional tone signal. The navigation task involves maneuvering the vehicle and approaching the platform, which is associated with the formation of zones with different acoustic visibility of the station emitters by the receiving elements of the vehicle. Three main zones of acoustic visibility are defined. The first zone is characterized by the reception of signals from all beacons and is the most informative. The second zone corresponds to a partial reduction in the number of observed beacons. The third zone (landing on the platform) is determined by a significant reduction in the number of observed beacons, and as a result, complicates the solution of the relative positioning problem. Taking into account the maneuvering of the AUV, the solution of the relative navigation problem should be performed in real time together with the execution of other tasks, which in turn imposes restrictions on the available resources of the computing module of the device. The paper presents a mathematical formulation of the relative navigation problem. A two-stage algorithm for finding a solution is proposed. At the first stage, the neighborhood of coordinates and orientation angles of the AUV is determined, at the second, the solution of the relative navigation problem is performed using the constraints and initial conditions obtained at the first stage. The paper proposes to perform the first stage based on a pre-trained neural network, describes the choice of network architecture, an approach to implementing the algorithm under a real-time OS, and presents the results of simulation modeling.