Sorokin A. Localization and navigation of mobile wheel robots using computational intelligence methods

In dissertation work the solution of the actual scientific and practical task of developing hybrid methods of localization and navigation of mobile wheel robots with the use of fuzzy and neural network models that can be effectively used in environments with variable properties is proposed. The object of research in work is the processes of localization and navigation of mobile objects with the use of methods of computing intelligence. Methods of research in work are based on the use of: methods of localization of moving objects in using radio signals, methods of optimizing routes in graphs, Kalman and Bayes filtration methods, methods of fuzzy logic, methods of machine learning with reinforcements and methods of neural network modeling. The paper proposes a method of localization and navigation of MPs in a medium with variable properties under the conditions of limited possibilities for remote control, which provides for the possibility of switching the mode of control of the robot to the state of autonomous navigation. The method of autonomous navigation of MR in an unknown environment with combined application of fuzzy model and RL-methods is proposed, which allows to improve the set of fuzzy rules using reinforcement signals. The improvement of the method of localization of mobile objects using iBeacon technology and accelerometer data in space with known maps of premises is proposed, which allows to reduce the amount of transmitters necessary for localization. The improvement of the method of localization of AMP with the use of Kalman filtration and Bayes filtration procedures, which allows taking into account the features of localization of the robot's position in the presence or absence of proprioceptive data, is proposed. The improvement of the method of autonomous navigation of MR in an unknown environment with obstacle bypassing with the use of fuzzy model, which allows to realize different types of behavior of the robot in the presence and absence of proprioceptive information. In this work, the method for determining the route of mobile objects movement using the modified Jump Point Search algorithm has been further developed, which allows reducing the number of computations required in comparison with the basic algorithms. The practical value of the work is that the obtained theoretical results are realized with the help of new intelligent procedures of localization and navigation of mobile objects, which ensures the possibility of their effective use in control systems of the MR in a medium with variable properties under the conditions of limited possibilities for remote control. The results of the testing of the proposed methods and the corresponding computational procedures confirm its efficiency and prospects of practical application. Some provisions, conclusions and recommendations of the dissertation work are used in the educational process for the preparation and modification of the course "Embedded Systems in Intellectual Robotics", as well as in course and graduation designing at the Department of Electronic Computing Machines of the Kharkiv National University of Radio Electronics, which is confirmed by the relevant act.