Currently, the field of underwater robotics is actively developing. The scope of tasks performed by autonomous underwater vehicles (AUVs) is expanding, and hence, the requirements for their autonomy are growing. These factors inevitably increase the time and cost of designing control and navigation systems for AUVs; in this connection, mathematical simulation begins to play an increasingly important role. This paper proposes a method for designing a control system for AUVs based on reduced-order models formed as a result of numerical simulation. The novelty of the work is that the AUV motion dynamics is represented as modified nonlinear transfer functions with nonlinear time-varying parameters, which are supposed to be determined from the results of numerical simulation. The proposed approach makes it possible to decompose the problem of control algorithm design, reducing it to an optimization problem, taking into consideration the cross-effect of the control loops. This can cause difficulties in the case that traditional analytical models are used. The implementation of the proposed approach is described on the example of designing an algorithm for AUV control in the vertical plane when it moves at a specified distance from the seabed. The effectiveness of the method is con-firmed in the course of similar mathematical experiments conducted on numerical models.

Accurate attitude determination of unmanned aerial vehicles (UAVs) is crucial for autonomous navigation, particularly when relying solely on gyroscope, accelerometer, and magnetometer measurements without utilizing the Global Positioning System (GPS). Reinforcement learning (RL) has emerged as a promising artificial intelligence technique applicable across various domains. This research introduces a novel approach that leverages RL to enhance the performance of the extended Kalman filter (EKF) in attitude estimation. The proposed method depends of RL which uses the Q-learning model and policy to find best solution to adjust autonomously the measurement noise covariance matrix within the EKF. By establishing a reward mechanism that incentivizes actions minimizing the prediction error relative to true measurements, the RL dynamically optimizes the measurement noise covariance matrix. This innovative integration of RL and EKF, referred to as RL-EKF, has been implemented and tested. Results demonstrate that RL-EKF significantly outperforms the traditional EKF, yielding marked improvements in attitude estimation accuracy. The improvement ratios showed that selected method is very effective in the field of attitude estimation.

The paper presents an overview of neural network (NN) technologies applied in navigation support of marine surface vehicles. It considers the problems related to motion path generation, detection, classification, and avoiding collisions with approaching ships, as well as navigation data integration and cybersecurity issues.

Astronomical calibration is defined as determination of the constant attitude of digital cameras relative to the inertial measurement unit (IMU) using ground-based star observations. The first part of [1] focuses on the calibration of relative attitude of all cameras, which provides combination of their separate measurements into the observations of one virtual camera. The second part examines the calibration of virtual camera attitude relative to the IMU. The calibration is performed on a simple test bench, which is unable to set the accurate attitude of the calibrated unit relative to the stars. Therefore, the calibrated parameters are estimated by the attitude increments over a time period between the astronomical measurements. Formulas for calculating the attitude increments from the inertial and astronomical measurements and their covariance matrices have been obtained. It has been shown that not all calibration trajectories provide unambiguous estimation of the calibrated parameters based on the attitude increments. The necessary condition for selecting the required trajectories has been formulated. The formula for finding the error covariance matrix of IMU attitude by the astronomical measurements has been received. This formula considers the star configuration observed with the virtual camera, its calibration errors described in the first part of the paper, and calibration errors in its attitude relative to the IMU obtained in the second part. The RMS error of the attitude calibration is experimentally found to be about 6 arcsec.

Concern CSRI Elektropribor, JSC, has accumulated extensive experience in the development and application of electrostatic suspensions (ESS) of sensitive masses of various inertial sensors—gyroscopes and accelerometers. The article presents the schemes that form the basis of these devices. An attempt is made to generalize the principles of construction and search for unified approaches to the creation of inertial sensors using ESS.

This study proposes an autonomous navigation approach for the Pioneer P3-DX Autonomous Wheeled Robot (AWR) in environments containing both static and dynamic obstacles. The robot utilizes the Artificial Potential Field (APF) algorithm for path calculation, while a neural network aids in zone classification. Three ultrasonic sensors provide distance measurements for hazard assessment. These measurements, along with relative velocity and angle data, aid in identifying regions of elevated risk (Zone 1) and those of lesser risk (Zone 2). Upon hazard detection, fuzzy logic facilitates effective collision avoidance by adjusting wheel velocities. Simulation results conducted in MATLAB and V-REP demonstrate the approach’s efficacy in navigating diverse obstacles, showcasing its adaptability and resilience compared to alternative algorithms. This research introduces an innovative methodology for autonomous mobile robot navigation, emphasizing its reliability and efficiency in traversing intricate environments with varying risk levels.

Astronomical calibration refers to determination of the constant mutual attitude of digital cameras and the inertial measurement unit using ground-based star observations. The first part of this paper considers calibrating the attitude of all cameras relative to the selected one. The calibrated vector contains the useful parameters (three attitude angles of each camera relative to the selected one) and interfering parameters (three attitude angles of the selected camera in the Earth-fixed frame at the time of shooting each frame). The system of nonlinear equations for determining the calibrated vector is based on the differences in the coordinates of images of the identified stars, which are calculated in the image planes of different cameras for each frame and then combined into a common residual vector. Atmospheric refraction and velocity aberration of light are considered when projecting the identified stars from the star catalog onto the image plane. Before solving the system of equations, intrinsic parameters of each camera are determined. Calibrating the relative attitude of real cameras provides a virtual camera with an extended field of view, which significantly reduces the error in star-based attitude determination. A virtual camera error model is provided that takes into account the errors in astronomical calibration. The results from experimental verification show that the error in astronomical calibration of the cameras’ relative attitude does not exceed 2 arcsec for each useful parameter.

The paper presents the results from the experimental study of a modified artificial neural network MFNN (minimum fuel neural network). Sparse representation of complex data with overcomplete basis and L0/L1 norm optimization are used instead of the classical fast Fourier transform (FFT) algorithm. The results showed a significant enhancement in the ability of obstacle recognition and autonomous railway control systems to distinguish between close objects such as trains on adjacent tracks at marshalling yards.

The paper presents the analytical review of an inertial measuring unit (IMU) rotation as a method to improve the accuracy of a strapdown inertial navigation system (SINS). There are two ways to improve the accuracy. One of them is based on the transformation of the error change pattern in the inertial sensors (IS) when using the IMU self-compensation rotation (SCR). The criteria for selecting an efficient SCR law to minimize the accumulated error in the parameters generated by SINS are presented. Along with the advantages of this technology, its weak points that may limit significantly the potentially achievable IMU accuracy are described. The other technique consists in increasing the observability of the IS error model components due to the IMU rotation while filtering the SINS errors. The IS error model is described, and the problem of recursive filtering of the SINS errors is stated to refine these errors, with the reference data on coordinates and motion velocity being available. The methods for quantifying the observability of the IS error model components are presented.

Under the water, the data are transmitted for ranges above dozens of meters using the acoustic modems. The acoustic modems are employed, in particular, for data transmission in positioning and navigation of both single and grouped underwater vehicles. Network communication methods are widely used for exchanging the navigation data between grouped autonomous underwater vehicles (AUVs). They are applied, for example, to keep the AUV formation geometry and/or to exchange the navigation data with the lead vessels. These methods can be combined, i.e., provide not only the exchange of mutual/relative positions between the AUVs for formation keeping, but also underwater georeferenced positioning for the group navigation in the geographical frame. A number of factors should be taken into account in the coordinated or cooperated AUV group navigation. As the number of interacting AUVs increases, their frequent addressing the acoustic medium is undesirable, because low capacity of the acoustic channel severely limits the underwater data traffic. Moreover, it is important to minimize the effect produced by the AUV acoustic system on the acoustic medium, particularly, to make the AUV less visible when its acoustic communication and positioning system is operating.