A nonlinear estimator concept for active vehicle suspension control

Abstract

A new Kalman filter based signal estimation concept for active vehicle suspension control is presented in this paper considering the nonlinear damper characteristic of a vehicle suspension setup. The application of a multiobjective genetic optimization algorithm for the tuning of the estimator shows that three parallel Kalman filters enhance the estimation performance for the variables of interest (states, dynamic wheel load and road profile). The Kalman filter structure is validated in simulations and on a testrig for an active suspension configuration using measurements of real road profiles as disturbance input. The advantages of the concept are its low computational effort compared to Extended or Unscented Kalman filters and its good estimation accuracy despite the presence of nonlinearities in the suspension setup.

Modified optimal control of a nonlinear active suspension system

“…By applying two Kalman filters in parallel 5 this estimator 5 In [11], the third one is necessary only for the estimation of the road excitation. Fig.…”

“…4. Pareto-front obtained from the 3-dimensional optimization structure achieves a remarkable estimation performance [11]. The nonlinearity of the damper is taken into account by considering the damper force as fictitious input signal, which can be calculated from the estimated damper velocity and the associated nonlinear characteristic.…”

“…For the implementation of the controller at the test rig, the nonlinear estimator concept presented by the authors in [11] is used to infer the state x of the quarter-vehicle system from the available measurement signals y m = [ẍ c ,ẍ w , x c −x w ], that represent a sensor configuration of modern production cars. By applying two Kalman filters in parallel 5 this estimator 5 In [11], the third one is necessary only for the estimation of the road excitation.…”

“…To achieve the best possible performance with respect to the conflicting control objectives, a multiobjective genetic optimization algorithm is utilized. The vehicle's vertical dynamic states are estimated from the available sensor signals using a nonlinear Kalman filter concept recently presented by the authors in [11].…”

Nonlinear State Estimation with an Extended FMI 2.0 Co-Simulation Interface

“…The corresponding implementation in Modelica is shown in Figure 9. The model consists of the two masses mass_body and mass_wheel, a linear spring damper component, which approximates the wheel behavior, a road model as explained in [18] or [22] and the body_spring and body_damper. As the motion of these two components are connected to the wheel and body motion by a push rodrocker kinematic (compare Figure 8 -left), the motion and the force of these components is scaled by a transmission ratio.…”