Realization of a Database-Driven Control System Using a CMAC

Abstract

As a nonlinear control algorithm, database-driven PID control (DD-PID) approach has been proposed to learn PID parameters based on a database. This method is based on a strategy in which PID parameters are determined based on neighboring data extracted based on the similarity between the query (current input/output data) and the information vector contained in the database. Since sorting operation is required in extracting the neighbor data, it is impossible to finish the calculation within a certain sampling interval for systems with fast response time, which is one of the hindrances in industrial applications. In addition, the DD-PID requires a large amount of storage memory in the database in order to obtain the desired control performance. On the other hand, one of the neural networks is the cerebellar model articulation controller (CMAC). It is a table-referenced adaptive learning controller. The major advantage of this method lies in the reduction of memory and computational load. This paper discusses a realization of the DD-PID by effectively utilizing the advantage of the CMAC.