Automation of linear and nonlinear control systems design by evolutionary compuatation

This paper develops a uniform definition of control system design problems using a vector space. Design difficulties by conventional analytical and numerical means are discussed. Associated problem-classification is presented, togetherwith a possible automation technique that transforms unsolvable design problems to solvable analysis and non-NP design problems. Practically realisable design automation methodology is then developed, using genetic algorithm based evolution programs to transform the non-NP problems to NP-complete problems. This methodology is applied to the establishment of uniform linear control systems. Design automation of linear and nonlinear control systems is illustrated through two examples. It is shown that this methodology leads to globally optimised control systems, in addition to the ease of design achieved by the computerised automation