Search

The aim of this work is to present a novel accurate estimator for vehicle dynamics. Following the multibody dynamics approach, a vehicle can be modeled with a high level of detail including non-linear dynamics. As a consequence, a rich simulation data-set is available for engineering analysis, richer than with vehicle analytical models. The proposed novel estimator is a new form of a dual Kalman filter. The first filter uses an indirect extended Kalman filter (i.e. the errorEKF) incorporating force estimation and using a vehicle multibody model. The second filter is an unscented Kalman filter (UKF) used to increase the accuracy of the errorEKF by estimating uncertain modeling parameters such as the mass of the vehicle and the tire-road friction coefficient. The performance of the proposed state-parameter-input (SPI) observer is tested in a simulation environment. The performance of the observer is demonstrated using two maneuvers, out of which one covers aggressive driving conditions. The results show that the new observer estimates with high accuracy the variables of interest for vehicle dynamics, such as the tire forces. [ABSTRACT FROM AUTHOR]

• A robust SMO is introduced to improve the capability when uncertainty are involved in the vehicle dynamic model. • The coupling effect between sensors drift and vehicle motion is studied, and the measurement error is compensated by the modified ENN. • The extended SCKF employs measurement from different sensors by a fusion strategy to deal with the severe driving motions. • Simulation and field tests with different maneuvers are implemented. To improve the safety and stability of land vehicles, this paper explores the estimation problem for vehicle states, including lateral velocity and attitude. First, a robust sliding mode observer is introduced to improve the adaptability for uncertain inputs, especially for the varying parameters in the vehicle dynamic model and longitudinal velocity. Furthermore, theoretical studies are performed to enhance the capability of the observer. In order to mitigate errors with the integrated navigation system, sensor drift model is primarily established based on a modified Elman neural network, so as to investigate the coupling between driving motion and errors. In addition, an extended square-root cubature Kalman filter is proposed to combine measurements from different sensors, utilizing a fusion strategy, to deal with severe driving motion and state estimation problems. Finally, simulation and field tests are carried out under a variety of maneuvers and conditions. The approach is compared with existing methods and evaluated experimentally, which indicates its effectiveness in improving the accuracy of vehicle state estimation. [ABSTRACT FROM AUTHOR]

• Robust and accurate road profile estimation method using smartphone data. • HC vehicle model identification through the automatic processing of smartphone data. • The influences of drive speed differences and changes are small. • The influences of differences in the left and the right paths are clarified. • The differences in drive speeds and vehicle dynamics are accurately compensated. This paper proposes a robust road profile estimation method and vehicle parameter identification method through an optimization with an objective function and constraint conditions on estimated profiles. The methods require only vehicle response measurements, enabling easy and inexpensive, yet effective, road condition monitoring through the automated processing of smartphone data. A half-car (HC) model, representing both bouncing and pitching motions, is employed for the profile estimation. The road profiles at the front and rear tire locations are included as state variables in the augmented state vector and are estimated by combining the augmented Kalman filter (AKF), Robbins–Monro (RM) algorithm, and Rauch–Tung–Striebel (RTS) smoothing. The two independent state variables, however, correspond to a single physical profile, while their distance coordinates differ by the wheelbase. Therefore, the vehicle parameters are optimized through the minimization of the difference between the identified road profiles at the front and rear tire locations using a genetic algorithm. Three objective functions and three constraint conditions are proposed to automatically select the best vehicle parameters. With this HC model, the road profile is subsequently estimated by combining the AKF, RM, and RTS methods. Through numerical simulations, the accuracy of the profile estimation and validity of the parameter identification are clarified. The influences of different drive speeds and difference between the left and the right profiles are numerically investigated. Drive tests with three different vehicles and a reference laser profiler show that the algorithm can automatically compensate for differences among vehicles with different drive speeds and estimate profiles accurately. [ABSTRACT FROM AUTHOR]