Evolution of Controllers under a Generalized Structure Encoding/Decoding Scheme with Application to Magnetic Levitation System
Bin Xin,Yipeng Wang,Wenchao Xue,Tao Cai,Zhun Fan,Jiaoyang Zhan,Jie Chen
DOI: https://doi.org/10.1109/tie.2021.3114700
IF: 7.7
2022-01-01
IEEE Transactions on Industrial Electronics
Abstract:Evolutionary search has been widely implemented for the adjustment of controllers' parameters. Nevertheless, the structure of controllers, which has a more important role in control systems, has been seldom studied. To this end, an evolutionary design method of controllers is proposed to optimize both structures and parameters simultaneously in this article. A controller is made up of a combination of some basic controller components and relevant parameters. The design of controllers can be transformed into an optimization problem involving the structure (represented by discrete vectors) and parameters (represented by real numbers). A generalized structure encoding/decoding scheme is developed. Guided by the performance indicators, intelligent algorithms for both combinatorial and numerical optimization are employed to iteratively and cooperatively evolve the controller structure and parameters, respectively. In order to effectively reduce some redundant or infeasible solutions, a set of generation rules for the controller structure are put forward, which also ensures the feasibility of the structure. Furthermore, this method is applied to a magnetic levitation ball system with nonlinear dynamics and external disturbance. Both simulation and experiment results demonstrate the effectiveness and practicability of the proposed method.