Learning to control: a heterogeneous approach

A heterogeneous approach to the design of an intelligent control system is presented. The purpose of such a system is to recreate the human intelligence within a controller. Application areas include autonomous control vehicles, robotics, and power plant controllers. First, the authors analyze the characteristics of the knowledge acquisition process and show that it is an evolutionary process. They further analyze the components of this process, which are then used as a guide to develop an intelligent control system. This analysis shows that the process may consist of a heterogeneous paradigm consisting of symbolic, fuzzy, and connectionist computation models. The knowledge structure and the learning/evolving mechanism are presented in detail. Finally, a program which learns to perform parallel parking is used as an example to show the effectiveness of this approach