In vehicle ego-motion estimation, vehicle control, and advanced driver assist systems the vehicle dynamics are described by a few key parameters. The side-slip gradient, being one of them, is used to model the lateral behavior of the vehicle. This parameter is rarely known precisely, since it depends on the vehicle’s mass distribution, its tires, and even the chassis setup. Thus, an online-estimation of the side-slip gradient is beneficial, especially in serial applications. Estimating the side-slip gradient with conventional vehicle sensors such as wheel-speed, steering, and inertial sensors poses a significant challenge since considerable dynamic excitation of the vehicle is required, which is uncommon in normal driving. Here, radar sensors open new opportunities in the estimation of such vehicle dynamics parameters since they allow for an instantaneous measurement of the lateral velocity. The use of radar for lateral velocity and parameter estimation has been proposed in literature before. However, radar on its own has proven to be unreliable and lacks both availability and accuracy. We present a novel and robust approach to estimate both the side-slip gradient and the lateral velocity by integrating radar-doppler measurements into a vehicle motion observer. The algorithm only requires low-dynamic, steady-state maneuvers. The solution is based on an adaptive Kalman-Filter assuring high accuracy and stability, while the number of radar sensors can be chosen arbitrarily. The algorithm has shown to estimate the side-slip gradient within 10% of its true value. It also rejects radar outliers and does not depend on permanent availability of the radar sensors. The approach requires little tuning which makes it applicable to mass-produced vehicles.
Session:
Vehicle Technology II
|
| 16:30 - 17:00