ANALYSIS OF THE SPEED OF THE DEFUZZIFIER IN THE TASK OF CONTROLLING THE ROBOT-MANIPULATOR

Abstract

Linear, nonlinear, modified, high-speed defuzzifiers based on the area ratio method are presented in this paper. The proposed defuzzifiers are used in a fuzzy digital filter device and make it possible to ensure the additivity of the robotic manipulator control system, since traditional models do not have this property. The essence of this development is to find a crisp value of the output fuzzy variable, which in this case are the regulation coefficients of the fuzzy digital filter. Reducing the number of computational operations provides an increase in the performance of the defuzzifier. The reduction in number of computational operations is carried out by eliminating the output variable’s truncated term’s height calculation, thereby reducing the computation time. A simulation model which was implemented in the MatLab Simulink system, for a neuro-fuzzy device of the robotic arm using linear and non-linear defuzzifiers is presented. The dependence of time graphs on the angle of rotation of the joints of the robotic manipulator are compared, based on the traditional center of gravity method and the method shown in this paper. It was found that the traditional center of gravity method does not ensure the fulfillment of the specified rotation angles of the links of the robotic arm, while the proposed models of defuzzifiers have this property, which can be seen from the presented dependency graphs. The simulation model of the device was also designed as a parallel-conveyor device for implementation in the field-programmable gate array of the Xilinx Spartan 3Е family. The analysis showed that the calculation time for a crisp value with high-speed defuzzification is 130 ns, which is two orders of magnitude higher than existing models. The experiment was conducted at a frequency of 100 MHz.