Application of fuzzy neural network for solution of kinematic problems of multilink manipulators in real-time control systems

Ganin P. E.

This paper examines the question of the application of fuzzy neural networks for solving inverse kinematics of multilink manipulators in real-time control systems. To ensure the required characteristics of the control system, a hybrid algorithm for finding solutions to kinematic problems has been developed. The developed algorithm for solving the inverse kinematics allows constructing the control system to work in real time with the control of the accuracy of calculations. The search for the solution of kinematic problems is based on the neural network approach in combination with numerical refinement algorithms. The paper presents the results of the study of the influence of neural network parameters on the quality of the solution in the developed search algorithm. The results of experimental studies shows the possibility of using the developed algorithm to find solutions to kinematic problems in real-time control systems.

Keywords:inverse kinematics, real-time control system, neural networks, iterative refinement