Integrating Model and Data-Driven Self-Organizing Control for Uncertain Manipulator

Higher control accuracy and a faster response speed while tracking a predefined reference path are critical for any industrial manipulator. However, how to use online data for compensation control based on imprecise dynamic models to improve tracking accuracy and robustness remains a challenging problem. In this article, we present a self-organized type-2 fuzzy feedback neural network control framework with a model and data fusion driven mechanism for tracking control of an uncertain manipulator. First, an internal feedback link of fuzzy neural network (FNN) is proposed to establish the relationship between the input and output of the FNN, which not only enhances the control accuracy but also accelerates the response speed by optimizing the network structure. Second, a kind of mixed index function is proposed for the rule self-organizing process in conjunction with the control framework, which can balance generational rules and each rule with various data in the current control period. Finally, to achieve control accuracy with limited real-time control data, a kind of input data reprocessing method in a self-organizing control framework is proposed. The theoretical analysis proves the stability and computational efficiency of the proposed control framework. Meanwhile, its effectiveness is verified by the simulation and experimental results of the manipulator.