Section: New Results
Zero-parameter mono and multi objective methods for the tuning of controllers
The synthesis of controllers for any kind of system is the main point in Automatic Control. The traditional approach is to use a simplified model of the system to control and/or use some reformulations of the specifications to tune an often efficient but suboptimal controller. In a more and more competitive industrial context, the design of high performances controllers has emerged as a crucial point to enhance the global productivity. However, the design of optimal controllers supposes the solution of non-convex and non-differentiable optimization problems, for which deterministic and (often) local search algorithm fail in the solution. In this work, Particle Swarm Optimization is used to solve the problem, and tested to define some controllers for a magnetic levitation. The use of standard settings and penalization terms leads to a zero-parameter and reformulation free method. Results are much than satisfactory and show that Evolutionary Computation could be of great interest in the Automatic Control field.