Full Vehicle State Estimation Using a Holistic Corner-based Approach

Abstract

Vehicles' active safety systems use different sensors, vehicle states, and actuators, along with an advanced control algorithm, to assist drivers and to maintain the dynamics of a vehicle within a desired safe range in case of instability in vehicle motion. Therefore, recent developments in such vehicle stability control and autonomous driving systems have led to substantial interest in reliable road angle and vehicle states (tire forces and vehicle velocities) estimation. Advances in applications of sensor technologies, sensor fusion, and cooperative estimation in intelligent transportation systems facilitate reliable and robust estimation of vehicle states and road angles. In this direction, developing a flexible and reliable estimation structure at a reasonable cost to operate the available sensor data for the proper functioning of active safety systems in current vehicles is a preeminent objective of the car manufacturers in dealing with the technological changes in the automotive industry.

This thesis presents a novel generic integrated tire force and velocity estimation system at each corner to monitor tire capacities and slip condition individually and to address road uncertainty issues in the current model-based vehicle state estimators. Tire force estimators are developed using computationally efficient nonlinear and Kalman-based observers and common measurements in production vehicles. The stability and performance of the time-varying estimators are explored and it is shown that the developed integrated structure is robust to model uncertainties including tire properties, inflation pressure, and effective rolling radius, does not need tire parameters and road friction information, and can transfer from one car to another.

The main challenges for velocity estimation are the lack of knowledge of road friction in the model-based methods and accumulated error in kinematic-based approaches. To tackle these issues, the lumped LuGre tire model is integrated with the vehicle kinematics in this research. It is shown that the proposed generic corner-based estimator reduces the number of required tire parameters significantly and does not require knowledge of the road friction. The stability and performance of the time-varying velocity estimators are studied and the sensitivity of the observers' stability to the model parameter changes is discussed. The proposed velocity estimators are validated in simulations and road experiments with two vehicles in several maneuvers with various driveline configurations on roads with different friction conditions. The simulation and experimental results substantiate the accuracy and robustness of the state estimators for even harsh maneuvers on surfaces with varying friction.

A corner-based lateral state estimation is also developed for conventional cars application independent of the wheel torques. This approach utilizes variable weighted axles' estimates and high slip detection modules to deal with uncertainties associated with longitudinal forces in large steering. Therefore, the output of the lateral estimator is not altered by the longitudinal force effect and its performance is not compromised. A method for road classification is also investigated utilizing the vehicle lateral response in diverse maneuvers.

Moreover, the designed estimation structure is shown to work with various driveline configurations such as front, rear, or all-wheel drive and can be easily reconfigured to operate with different vehicles and control systems' actuator configurations such as differential braking, torque vectoring, or their combinations on the front or rear axles. This research has resulted in two US pending patents on vehicle speed estimation and sensor fault diagnosis and successful transfer of these patents to industry.