Co-design of controller and setup configuration using Genetic Algorithm

Abstract

In many structures the decision on how to apply actuators and sensors is a complicated puzzle. A balance between implementation cost and achievable performance must be found, and this proves to be a challenging task. In this paper, an optimization procedure is proposed to co-design the number of actuators and sensors on the one hand and simultaneously determine the corresponding optimal controller feedback gains on the other hand. Both are optimized to obtain optimal control performance. Starting from a state-space representation, the presence or absence of actuators and sensors is described as selection binaries. Furthermore, many non-linearities are present as for example the maximum control effort u or implementation cost change discontinuously when different configurations are used. A proposed method to answer these problems is to use a novel Genetic Algorithm implementation. This way, a resulting optimization procedure is formulated to define the optimal hardware configuration choosing from several possible actuator types on the one hand. On the other, it can concurrently determine the feedback gains that make optimal use of the available maximum actuator control effort u.