Simulation-based multi-objective optimization of a fuzzy controller for semi-active suspension
This paper presents a simulation-based design process for a fuzzy-controlled semi-active suspension system applied to a real-time vehicle simulation. Firstly a control architecture for the fuzzy-logic controller is defined. Secondly different Genetic Algorithms (GAs) are configured for the optimization of the control parameters. A multi-objective optimization is performed in order to simultaneously improve safety and comfort of the driving vehicle, defined using two independent cost functions. The contact forces at the wheels estimate the driving safety and the vibration in the car body indicates driving comfort. The performance of the controller is com-pared for each set of parameters obtained by the different GAs adopted. A reduced order real-time simulation environment has been set up for a holistic vehicle simulation, which includes a Finite Element Model (FEM) of the chassis, non-linear suspensions, multi-body physics and the designed digital controller. Finally, the real-time environment is integrated into an overall optimization process and is used for fitness evaluation. The semi-active system under test allows shifting the Pareto front beyond the limit of passive systems, achieving simultaneously better safety and comfort. The different GAs come up with various near-optimal solutions, which are compared using their Pareto fronts.