Design of Robust Observers for Active Roll Control

Abstract

This paper presents robust observers for active roll control. Generally, active roll control adopts feedback control structure with roll angle and roll rate signals which should be measured with a sensor. For active roll control, it is necessary to estimate the roll angle with roll rate measurement using an observer because the roll angle is difficult to measure, compared to the roll rate. However, an observer designed with fixed parameters can give poor estimation performance since there are large uncertainty in parameters of real vehicles. A robust observer is designed to cope with the parameter uncertainty. In this paper, a dual Kalman filter and Kalman filter with constant velocity model are adopted as a robust observer. The performance of the proposed algorithm has been investigated via computer simulations and vehicle tests.

Active Roll Stabilization With Disturbance Feedforward Control

“…Since CDC is a semi-active device, it can only prevent the compression and expansion of the suspension, and it cannot create active force. This fact is expressed by four conditions such as equations (25), (26), (27), and (28), and the value of the CDC current is determined accordingly [8][9][10]. In these equations, Fsr and Fsl are the forces of right and left suspensions, generated by the command currents, ir and il, of CDC, respectively.…”

“…For the given parameters, T and C, the state variables are estimated with DTKF. With the estimated state variables, the parameters, T and C, are updated with RLS [25]. Then, these estimated values are used to update Ad and Bd in (34), and these matrices are used in DTKF.…”

“…Recently, several studies have been conducted to estimate the roll and pitch angles using a low-cost IMU [23,24]. Following the previous works, a discrete-time Kalman filter (DTKF) is adopted to estimate the roll angle from measured roll rate signal [25].…”

“…Fig. 1 is the lateral acceleration and the roll rate signals measured using the accelerometer and the inertial measurement unit (IMU) installed on a real vehicle during the dual lane change steering at a speed of 60 km/h, respectively [22,25]. If these signals are used for control, the noise components in the signals are directly passed to the control input, which can generate severe chattering in roll responses.…”

“…It has been known that an extended Kalman filter (EKF) is equivalent to RLS. Following the idea, the parameters of a roll model were estimated with dual EKFs [25]. RLS is simple to design and easy to implement.…”