Known algorithms for detecting and matching singular image points are presented when solving the problem of autonomous positioning of an unmanned aerial vehicle (UAV) equipped with a video camera in the absence of a signal from the global satellite navigation system. The idea of autonomous UAV positioning was to sequentially compare adjacent pairs of video stream frames to track the movements of the video camera carrier in the time interval between the mentioned frames. The purpose of the study is to identify those that best allow tracking the coordinates of the carrier based on the results of a comparative analysis of algorithms. The real flight data used to construct the control trajectory was obtained using a mock-up model of an aircraft-type aircraft equipped with an on-board inertial navigation system with a set of microelectromechanical sensors and a visual odometer with a nadir-oriented video camera. The comparison of real flight data and the results of the algorithms was carried out according to several quality indicators: the minimum standard deviation of the positioning error, the maximum deviation from the control trajectory, and the average time required to process a pair of adjacent frames. The results of comparing the algorithms are summarized in a table, and graphs of the control trajectory and estimates obtained using these algorithms are also provided for visual comparison. According to the set of positioning quality indicators, the neural network algorithms d2net, darkfeat and XFeat(dense) are recommended for implementation.

A mathematical model of dynamic processes in a multi-chamber liquid rocket engine with a central body (liquid propellant rocket engine with a central heating element) is presented, mathematical models of a branched tube, a combustion chamber, a flow regulator, and the process of flowing around the central body with jet streams are described. Expressions using the components of the state vector are proposed for the non-stationary parameters of the model. The numerical values of the parameters are consistent with the results of thermodynamic calculations using the Navier— Stokes equations for the alcohol-oxygen fuel pair and the geometric parameters of the engine demonstration model. The thrust of a separate chamber and its corresponding component of interaction with the central body are expressed as functions of the components of the state vector of the engine. The thrust vector of the engine is represented as the sum of the thrust vectors of all chambers, adjusted for the interaction of gas flows with the central heating system. A control variant of a liquid propellant rocket engine with a central heating system is formed based on a predictive model, while the control characteristics are obtained by numerical modeling.

A routing algorithm designed for use in geographic networks is presented. The basis of the proposed procedure is Dijkstra's algorithm, used to find the shortest path in directed graphs. To increase the efficiency of message delivery in the network, an additional stage is proposed — the preliminary ordering of the router's output queue, which is based on optimal rules that affect the data transfer rate. The criterion of optimality in the proposed approach is the minimum message delivery time, which is a key factor in modern network technologies. The rules of message ordering are formulated for different scenarios when there are both partially ordered and unordered groups of messages in the queue, which makes it possible to adapt the algorithm to different network conditions and improve overall system performance. The modeling results are presented, demonstrating that the implementation of the proposed algorithm leads to a significant reduction in message delivery time compared with traditional routing methods. This opens up new opportunities for the development of high-performance network solutions that can effectively cope with the growing demands for speed and reliability of data transmission in modern information technology.

A dynamic model of the optical glide path axis stabilization system is presented. An algorithm for estimating the parameters of the system state vector based on an observer in the form of a Kalman filter and a control algorithm using a linear-quadratic regulator are constructed. The system is synthesized with nominal model parameters and perturbation characteristics that depend non-linearly on the state vector. It is shown that the stabilization error under these conditions is ±46". The sensitivity analysis of the constructed algorithm is carried out to account for the possible deviation of the model parameter values from the nominal values.

The design features and test results of samples of compensation-type pendulum accelerometers with torsion suspension are presented. A comparative analysis of the results of precision tests of the manufactured prototypes in two modes is carried out — independently and as part of an inertial module. The obtained characteristics of the devices and their comparison with typical values of similar characteristics of macromechanical accelerometers are presented.

The problem of choosing the modulation coefficient for the magnetometric system of a quantum rotation sensor in order to increase the signal-to-noise ratio during measurements is considered. Experimental results are presented that demonstrate the possibility of evaluating the quality of resonance in a quantum rotation sensor cell in two different ways: the classical laboratory method based on varying the modulation coefficient and analyzing the amplitude of the output signal, and the proposed method based on estimating the parameters of the magnetometric signal. There is also a comparison of the time spent on the implementation of each of the methods. Based on these studies, an approach has been developed to choose the value of the modulation coefficient, taking into account the current resonance quality in the system.

The methodology, models, and algorithms for calibrating the parameters of the instrumental and geometric errors of the odometer and the steering angle sensor of land-based wheeled vehicles, in particular agricultural ones, are presented. Odometric parameters are calibrated using readings from low-accuracy inertial sensors, as well as information from the global satellite navigation system signal receiver, if available. Odometric parameters include the error of the odometer scale factor, misalignment of the instrument axes of the free-form inertial navigation system and the "measuring" axis of the odometer along the heading and pitch channels, the offset of the zero signal and the error in the scale factor of the DOP coefficient of the steering mechanism of the machine. The calibration problem is considered for two variants of structural arrangements of agricultural machines: classical and articulated. The results of applying the methodology and algorithms in field experiments are presented, as well as the results of covariance analysis of the estimability of odometric parameters depending on the nature of the object's movement. The proposed approaches have been tested on a representative set of simulation and real data and are used on many agricultural machines: tractors, combines, self-propelled mowers and self-propelled sprayers.

A mathematical model of an unmanned agricultural platform is being built in the MatLab — Simulink environment, which takes into account kinematic and dynamic parameters that characterize the interaction of platform nodes with each other and with the surrounding terrain. A ready-made Simscape Mechanics package is used to describe the physical system. The possibility of using the constructed model to generate stochastic signals from micromechanical gyroscopes and accelerometers when the unmanned platform is moving along a selected area on the map is being investigated. The implementation of simulation modeling is obtained by converting an object assembly developed in CAD. The auxiliary role for conversion is performed by the ready-made Simscape Multibody Link plug-in installed in CAD. In addition to the model of the agricultural platform, an algorithm for modeling the forces of contact interaction of wheels with the ground is presented. The results of simulation and field experiments are presented, indicating the possibility of using the proposed approach to generate signals close to real gyroscopes and accelerometers. The ADIS 16480 and its technical parameters from the documentation were selected as the investigated inertial meter. The developed technology of simulation modeling of a real object can be expanded and used to verify algorithms for motion control of autonomous systems and navigation.