Learning based approach to the design of hierarchical hybrid fuzzy PID controllers

The development of a systematic procedure for the design of fuzzy logic controllers (FLCs) has been of interest since their introduction. As the dimension of the system's input/output space increases, the design problem becomes exponentially more difficult. By partitioning the input space into smaller, more manageable pieces, it is possible to simplify the design procedure. The problem then becomes one of integrating the regional controllers designed for each input space partition. This paper demonstrates a systematic approach to the design of a hierarchical controller through the application of a learning-based algorithm to integrate these regional controllers. The proposed method is applied to a hybrid fuzzy-PID controller operating on a two degree-of-freedom robot arm, and is shown through simulations to compare favorably to fixed coefficient PID control and variable structure control.