Genetic Adaptive and Supervisory Control

A genetic algorithm (GA) uses the principles of evolution, natural selection, and genetics to offer a method for parallel search of complex spaces. Motivated by past work using the GA as a tool for off-line computer-aided-design of control systems, we introduce several ways to use the GA for on-line identification and controller tuning. In particular, we introduce two new, but closely related approaches to genetic adaptive control which we call “genetic model based control” (GMBC) and “genetic model reference adaptive control” (GMRAC). In these techniques a GA manipulates a set (population) of parameterized controllers in order to evolve the controller most capable of providing good performance for the current plant operating conditions. Next, we introduce a hierarchical GA-based supervisory controller (GASC) that seeks to evolve the GMRAC’s fitness evaluation procedure so that the controllers chosen by GMRAC guide the system towards reduced error and decreased use of control energy. The use of GASC removes the need for the designer to specify GMRAC parameters. In addition, we show how GA-based estimation (GAE) can be used to evolve a model of the plant that is used in the model based controllers GMBC and GMRAC. A cargo ship steering problem is used throughout this paper as a theme example to illustrate each of the genetic adaptive control concepts and techniques.