Tovkach I. Methods of adaptive estimation of unmanned aerial vehicles movement parameters based on measurement of the sensor network.

An actual scientific problem of developing methods for adaptive estimation of the maneuvering UAV movement parameters based on TDOA- and RSS-measurements of the sensor network and their complex use that allow to increase the accuracy of determination its location has been solved in the dissertation. The first section of the dissertation shows the relevance of the task of determining of the UAV movement parameters based on the sensor network data, due to the emergence of a new class of threats using UAV, which leads to the need to develop systems that solve the problems of detection, location and UAV movement parameters. The analysis of methods for locating UAVs based on TDOA- and RSS- measurements of the sensor network based on mathematical methods of maximum likelihood and least squares is carried out. It is noted that in known methods of calculating the UAV coordinates is performed after arrival of measurements from all sensors. The methods of Kalman filtering and adaptive estimation of object motion parameters are considered and conclusion that it is expedient to use the mathematical apparatus of mixed Markov processes in discrete time to solve the formulated scientific problem. The research problem was formulated. The second section of the dissertation, with use of a mathematical apparatus of the extended Kalman filtering, recurrent algorithms for determining the UAV location according to a sensor network data were developed on the basis of: TDOA-measurements; RSS-measurements; complex processing of TDOA- and RSS- measurements, which, after formation of initial conditions based on measurements of the minimum number sensors, allow recurrently to specify at each step l the UAV location in process of arrival of measurements from other sensors. With the help of statistical modeling, the analysis of the accuracy characteristics of the developed algorithms is performed. Comparing with lower bound of Cramer-Rao and the known algorithms is carried out them. The analysis of the influence of sensors configuration (topology) of the sensor network on the UAV location accuracy was also performed. Sensor network configurations have been obtained, which can be recommended in cases of known and unknown directions of UAV appearance. The third section of the dissertation, on the basis of a mathematical apparatus of the mixed Markov processes in discrete time optimal and quasioptimal algorithms of the adaptive filtering of the maneuvering UAV movement parameters according to a sensor network data are synthesized on a basis of: TDOA-measurements; RSS-measurements; complex processing of TDOA- and RSS- measurements. The optimal adaptive algorithms are recurrent and describe the evolution of a posteriori distributions of the estimated parameters, and the optimal devices are multi-channel and belong to the class of devices with feedbacks between channels. Quasioptimal algorithms obtained using polygous approximation of a posteriori distributions. The analysis of the obtained quasioptimal algorithms is carried out with the help of statistical computer simulation using the example of estimation of the UAV movement parameters performing in advance at random time intervals unknown types of maneuvers. The fourth section of the dissertation analyzes the effectiveness of algorithms developed in sections 2 and 3 in space, which corresponds to a real situation of estimating of the UAV movement parameters emitting a radio signal. The developed algorithms allow to obtain UAV location characteristics close to the potentially achievable ones determined by lower bound Cramer-Rao over all three spatial coordinates. Also developed adaptive filter based on TDOA- and RSS-measurements allow to recognize the hovering and almost uniform motion of the UAV with a probability close to unity. The adaptive filter based on complex processing of TDOA- and RSS-measurements allows to significantly increase the probability of recognition of short-time UAV maneuvers. In addition, a comparative analysis of the computational costs required for the implementation of synthesized algorithms, and also possibilities of their implementation on the basis of modern computer systems is carried out.