Open Conference Systems, CAR 2017

Font Size: 
Real-Time Parameter Estimation for Vehicle Systems
Corina Sandu

Last modified: 2017-09-05

Abstract


Many accidents still occur due to vehicle rollover. The efficacy of stability control system is highly dependent on the accuracy of the parameters used in establishing the states of the vehicle, and thus the propensity of the vehicle to roll over. It is very important to have correct and on-the-fly knowledge about critical vehicle characteristics that have a strong impact on handling, stability, and rollover propensity. This study focuses on developing advanced real-time parameter estimation techniques, and the appropriate vehicle models.

Parameter estimation, especially for real-time applications, is a challenging problem. Often, one needs to trade-off robustness for accuracy. For example, Lyapunov estimation techniques guarantee stability but do not guarantee parameter accuracy. It is very important to be able to observe the states of the system using sensors or observers. The research presented here significantly improves the Generalized Polynomial Chaos Extended Kalman Filter (gPC-EKF) for state-space systems; it also expands it for estimating parameters in linear regression problems.

Modeling of ground vehicles is a very complex problem, especially when the model has to be able to rely on input parameters collected on-the-fly by various sensors. Comprehensive multibody dynamics models represent the dynamics of the vehicle accurately, but the computational time and the extensive amount of data needed are too significant to be employed in real-time systems. The literature focuses primarily on vehicle models designed for control purposes or off-line vehicle dynamics simulations and are not suitable for parameter estimation.

In this study, in the search for the best vehicle model to be used for real-time parameter estimation, several vehicle models have been developed: a Load Transfer Model (LTM), a Modified Load Transfer Model (MLTM), and a High Frequency Load Transfer Model (HFLTM). The parameters in the model are updated using the gPC-EKF method. The mass and the horizontal center of gravity (CG) position of the vehicle are estimated to high accuracy.

Keywords


Parameter Estimation; System Identification; Extended Kalman Filter; Generalized Polynomial Chaos; Load Transfer; Rollover Propensity

Conference registration is required in order to view papers.