Enhancement of vehicle handling dynamics model using global searching algorithm and estimation theory

Computer simulations are popular in modeling vehicle system dynamics. However further refinement of the vehicle dynamic model is required for extensive use in the automotive industry. In this dissertation, the model refining procedure is illustrated by developing more reliable kinematic models verified with laboratory test results, instrument test data, and a mathematical optimization method. More specifically, simple kinematic models are developed for reduced computation times using ADAMS. They are tuned by the gradient-based optimization technique using the results from a laboratory testing facility, which includes the compliance effect in order to use the kinematic models in dynamic simulations. Finally these models are verified in the full vehicle multibody dynamic simulation with instrument measurement data from the lane change and slowly increasing steer handling testing. Also the estimation methods and the global searching algorithm are introduced for enhancement of the vehicle dynamic model. It is already known that the gradient-based optimization techniques used in the model refinement stage do not readily yield the optimum since handling maneuvers have nonlinear characteristics. The simplified models are developed through one of the global searching algorithms, i.e., the Genetic Algorithm. The refined models are used for estimating the state vector and tire forces during a lane change handling test using the extended Kalman filter and the observer theory; the Luenberger and Sliding Mode observer. Information of the vehicle state vector and tire forces is essential to determining the stability of vehicle dynamic behavior. The results from the estimation methods are compared with that of the full vehicle simulation for confirmation. Hence, the more reliable design of the vehicle steering and suspension system can be achieved using the confirmed information for the state vector and tire forces.