Adaptive Control Strategy for Automotive Magnetorheological Dampers Based on Artificial Neural Networks

In vehicle systems, the suspension is fundamental in achieving acceptable levels of ride comfort, while ensuring ride safety. Semi-active suspensions allow changing the suspension damping to prioritize different performance metrics. Multiple control techniques have been developed to find a suitable trade-off between comfort and road holding. This manuscript proposes the use of standard (proportional-integral-derivative) controllers tuned in real time by an artificial neural network. The formulation of the controller considers a magnetorheological damper represented through the Bouc-Wen model. A stochastic gradient descent algorithm with backward propagation is used to train the artificial neural network that then selects the controller gains in real time. This technique is tested numerically through quarter and full car models, with the latter one running on the automotive simulation software CarSim. The obtained results highlight significant improvements of the proposed approach in comparison to state-of-the-art controllers. Furthermore, the study proves the viability of running four controllers on a real-time embedded hardware platform through processor-in-the-loop tests.