Linear control of car suspension using nonlinear actuator control

In classic passive suspension design, comfort and road handling are conflicting design criteria. An active or semi-active suspension offers the possibility to vary the damper characteristics along with the road profile. Such systems have discrete settings or are limited in bandwidth up to 10Hz. This paper discusses the development of a controller for a passenger car using a recently developed active shock absorber, were two current controlled valves continuously vary the damper characteristics. The aim is to control the cars rigid body modes heave, roll and pitch, as well as wheelhop and the first bending and torsion modes. The control design is based on damper and car models that are valid up to these frequencies. The shock absorber, which is a highly nonlinear hydraulic system, is modelled by a dynamic neural network. Using the active shock absorber as an actuator, the car itself is identified using linear identification techniques based on random multisine excitations to find the best linear approximation of its dynamic behavior. Based on the identified linear car model, linear controllers are derived. These controllers calculate the actuator forces that have to be generated by the active shock absorber. Based on the nonlinear damper models, these desired forces are converted into appropriate currents to be applied at the damper valves. This requires a nonlinear model inversion. The paper presents experimental results of the proposed controllers on a quarter car test setup. The performance of several linear car controllers in combination with nonlinear damper model inversion strategies are compared.