UC Berkeley

This thesis develops vehicle parameter identification algorithms, and applies identified parameters to a controller designed for safe path following.

A tire-road friction coefficient is estimated using an in-tire accelerometer to measure acceleration signals directly from the tires.

The proposed algorithm first determines a tire-road contact patch with a radial acceleration profile.

The estimation algorithm is based on tire lateral deflections obtained from lateral acceleration measurements only inside the contact patch.

A new model is derived for the lateral deflection profiles, which provides robustness to orientation-variation of the accelerometer body frame during tire rotation.

A novel algorithm is developed to identify three inertial parameters: sprung mass, yaw moment of inertia, and longitudinal position of the center of gravity.

A correlation of inertial parameters is derived and is used for the identification algorithm.

Inertial parameters and vehicle states are simultaneously estimated with a dual unscented Kalman filter based on a nonlinear vehicle model.

In order to activate and de-activate different modes of the proposed

algorithm, a local observability analysis is performed with the nonlinear vehicle model.

The performance and robustness of the proposed approach are demonstrated with extensive CarSim simulations and experimental tests on a flat road with a constant tire-road friction coefficient.

Following a curved road can be dangerous if autonomous vehicles do not take roll motion into consideration.

A control algorithm is designed to prevent a dangerous vehicle state induced by roll motion while following a curved road.

Roll motion is suppressed throughout cornering with model predictive control.

A four-wheel nonlinear vehicle model with roll dynamics and a tire brush model are utilized for the prediction of the vehicle state.

An optimal balance in the trade-off between vehicle speed and

roll motion is achieved with full braking as a control actuator.

Identified vehicle inertial parameters are incorporated into the designed controller.

CarSim simulations illustrate the performance of the proposed controller and the effect of the vehicle parameter estimator.