Authors:
Elias Reichensdörfer
1
;
Dirk Odenthal
2
and
Dirk Wollherr
3
Affiliations:
1
BMW Group, Knorrstr. 147, 80788, Munich, Germany, Department of Electrical and Computer Engineering, Chair of Automatic Control Engineering, Technical University of Munich, Theresienstr. 90, 80333, Munich and Germany
;
2
BMW M GmbH, Daimlerstr. 19, 85748 Garching near Munich and Germany
;
3
Department of Electrical and Computer Engineering, Chair of Automatic Control Engineering, Technical University of Munich, Theresienstr. 90, 80333, Munich and Germany
Keyword(s):
Nonlinear Control Structure Design, Lyapunov Equation, Grammatical Evolution.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Computational Intelligence
;
Evolutionary Computation and Control
;
Evolutionary Computing
;
Genetic Algorithms
;
Industrial Engineering
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Nonlinear Signals and Systems
;
Performance Evaluation and Optimization
;
Robotics and Automation
;
Signal Processing, Sensors, Systems Modeling and Control
;
Soft Computing
;
Vehicle Control Applications
Abstract:
A new method for the automated synthesis of nonlinear control laws for nonlinear control systems using grammatical evolution is presented. The controller structure, its parameterization and a quadratic Lyapunov function are the result of a nonlinear, nonconvex optimization process. Evolutionary algorithms based on grammatical evolution are used to find candidates for the control law. These are evaluated using a fitness function incorporating eigenvalue specifications on the linearized closed loop system and bounds on the control input signals. The guaranteed domain of attraction subject to the closed loop performance and stability specifications is maximized by evaluating the solution of the Lyapunov equation on the nonlinear system. The method is tested on two different control systems that contain different types of nonlinearities. The results show that the proposed approach is capable of outperforming state of the art methods by providing stronger stability guarantees and/or bette
r closed loop performance while making less restrictive assumptions.
(More)