Vibration control of semi-active suspension by the neural network that learned the optimal preview control of MLD model

Due to developments in image processing technologies, it is now able to track road ahead contours while driving. We propose a new semi-active suspension control mechanism considering forward road surface form. One can express a vehicle model with a semi-active suspension as an MLD ( Mixed Logical Dynamical) model. Accurate measurement of the road ahead’s form guarantees the availability of information on future disturbances prior to the vehicle’s travel. In this work, we develop the finite time optimization problem of the MLD model in consideration of the future disturbances as a MIQP ( Mixed Integer Quadratic Programming) problem in the same way as the conventional optimal control problem without future disturbance. One can find the MIQP solution by means of a generally accessible solver program. But since the MIQP problem is NP-hard, it is difficult to find the control action inside the control cycle period needed by the vibration control problem with general computers for vehicles. In this work, we build an approximative function for the intended controller in order to reduce computational load. The approximation is using a multilayer neural network. A simulation was used to assess the proposed control strategy. With the same suspension stroke, the proposed method was able to attain better ride comfort in the simulation study than the conventional MLD predictive control and the Skyhook approximation approaches. Moreover, the suggested approach helps to create the control signal inside the control cycle period.