Model-free development of control systems for a multi-degree-of-freedom robot

Abstract

For a multi-degree-of-freedom (MDOF) robot, its dynamics model is very complex, and there are many contained terms. Moreover, with the increase of the degree-of-freedom (DOF), the number of the terms contained in the dynamics equation increases geometrically, and the dynamics equation has the characteristics of highly nonlinear and serious coupling, therefore, it is difficult to achieve the accurate and efficient control. In particular, when the uncertain factors such changes in the load, friction and disturbance are considered, the problems of control are more obvious. To deal with these problems, the two model-free intelligent control systems are designed in this paper: (1) The adaptive sliding mode control (ASMC) system; (2) the fuzzy neural network control (FNNC) system. For the ASMC, the consumed time is shorter, and the efficiency is higher, but the control accuracy is relatively poorer. However, for the FNNC, the control accuracy is relatively higher, but the consumed time is longer, and the efficiency is poorer. In order to give full play to the advantages of the two intelligent control systems, the ASMC and the FNNC are combined to form the adaptive sliding mode-fuzzy neural network control (ASM-FNNC) system, which the priority is given to the ASMC, and the error thresholds are set, when the control error exceeds the thresholds, switch to the FNNC. Finally, the proposed control scheme is applied to a six DOF robot, to verify its effectiveness.