ABSTRACT
In this paper, recent research results [1] are presented which demonstrate the effectiveness of genetic algorithms in the control of dynamic systems. Genetic algorithms are search algorithms based upon the mechanics of natural genetics. They combine a survival-of-the-fittest among string structures with a structured, yet randomized, information exchange to form a search algorithm with some of the innovative flair of human search. While randomized, genetic algorithms are no simple random walk. They efficiently exploit historical information to speculate on new search points with improved performance.
Two applications of genetic algorithms are considered. In the first, a tripartite genetic algorithm is applied to a parameter optimization problem, the optimization of a serial natural gas pipeline with 10 compressor stations. While solvable by other methods (dynamic programming, gradient search, etc.) the problem is interesting as a straightforward engineering application of genetic algorithms. Furthermore, a surprisingly small number of function evaluations are required (relative to the size of the discretized search space) to achieve near-optimal performance.
In the second application, a genetic algorithm is used as the fundamental learning algorithm in a more complete rule learning system called a learning classifier system. The learning system combines a complete string rule and message system, an apportionment of credit algorithm modeled after a competitive service economy, and a genetic algorithm to form a system which continually evaluates its present rules while forming new, possibly better, rules from the bits and pieces of the old. In an application to the control of a natural gas pipeline, the learning system is trained to control the pipeline under normal winter and summer conditions. It is also trained to detect the presence or absence of a leak with increasing accuracy.
