Estimation with Applications for Automobile Dead Reckoning and Control
This dissertation focuses upon parameter estimation and how parameterized vehicle models may be used for navigation and stability control. It begins by developing a nonlinear estimation scheme which yields extremely consistent parameter estimates in simulation and experiment. This estimation scheme is then used to experimentally identify the sensitivity of the tire parameters, longitudinal stiffness and effective radius for two tires under several different driving conditions. The data clearly show that there are several important parameters which govern tire longitudinal stiffness behavior in the low slip region. At a minimum, inflation pressure, tread depth, normal loading and temperature have a strong influence on longitudinal stiffness estimates; the change from dry to wet asphalt had the smallest effect on longitudinal stiffness estimates.
The work then moves onto parameter estimation for a vehicle navigation filter which uses differential wheelspeed measurements to estimate vehicle yaw rate for dead reckoning during GPS unavailability. The new navigation filter is compared with a similar filter which uses an automotive grade gyroscope to measure yaw rate. Test results show the gyro and wheelspeed based schemes perform equally well when navigating smooth road surfaces. However, the wheelspeed-heading estimator's position errors grow about twice as fast as the gyro based system's when navigating speed bumps and uneven road surfaces. Contrary to previous work in the literature, it is shown that dead reckoning position error is not a function of encoder quantization error and that longitudinal slip of the vehicle tires is not a dominant error source. The limiting factor for dead reckoning performance is road surface unevenness. The thesis concludes by applying a nonlinear controller design methodology to vehicle stability control which is enabled by parameterized vehicle models. As an example, a controller which prevents untripped vehicle rollover is designed. Numerical simulations on a nonlinear vehicle model show that the designed control laws effectively track the driver commands during extreme maneuvers while also maintaining a safe roll angle. During ordinary driving, the controlled vehicle behaves identically to an ordinary vehicle.