Skip to content Skip to navigation

State and Parameter Estimation for Vehicle Dynamics Control Using GPS

January, 2004

Many types of vehicle control systems can conceivably be developed to help drivers maintain stability, avoid roll-over, and customize handling characteristics. A lack of state and parameter information, however, presents a major obstacle. This dissertation presents state and parameter estimation methods using the Global Positioning System (GPS) for vehicle dynamics control. It begins by explaining basic vehicle dynamic models which are commonly used for vehicle dynamics control. The dissertation then demonstrates a method of estimating several key vehicle states – sideslip angle, longitudinal velocity, roll and grade – by combining automotive grade inertial sensors with a GPS receiver. Kinematic estimators that are independent of uncertain vehicle parameters integrate the inertial sensors with GPS to provide high update estimates of the vehicle states and the sensor biases. With a two-antenna GPS system, the effects of pitch and roll on the measurements can be quantified and are demonstrated to be quite significant in sideslip angle estimation. Using the same GPS system, a new method that compensates for roll and pitch effects is developed to improve the accuracy of the vehicle state and sensor bias estimates. In addition, calibration procedures for the sensitivity and cross-coupling of inertial sensors are provided to reduce measurement error further.
To verify that the estimation scheme provides appropriate estimates of the vehicle states, this dissertation shows that the state estimates from real experiments compare well with the results from calibrated models. The performance of the estimation scheme is also verified by statistical analysis. Results from the statistical analysis match predictions from the kinematic estimators. Since the proposed estimation scheme is based on a cascade estimator structure, the convergence of the cascade estimator structure is also proven. As an application, the estimated vehicle states are used to virtually modify a vehicle’s handling characteristics through a full state feedback controller. Results from this application show that the estimated states are accurate and clean enough to be used in vehicle dynamics control systems without additional filtering.
The dissertation then examines parameter estimation for vehicle dynamic models. Several important vehicle parameters such as tire cornering stiffness, understeer gradient, and roll stiffness, can be estimated using the estimated vehicle states. Experimental results show that the parameter estimates from proposed methods converge to the known values. Finally, the dissertation presents a new method for identifying road bank and suspension roll separately using a disturbance observer and a vehicle dynamic model. Based on the estimated vehicle parameters, a dynamic model, which includes suspension roll as a state and road bank as a disturbance, is first introduced. A disturbance observer is then implemented from the vehicle model using estimated vehicle states. Experimental results verify that the estimation scheme gives separate estimates of the suspension roll and road bank angles. The results of this work can improve the performance of stability control systems and enable a number of future systems.