This dissertation presents a study on lateral motion control of an electronic-four-wheel drive (e-4WD) vehicle based on vehicle state estimation making use of easily available in-vehicle sensors and a low-cost standalone global positioning system (GPS). In keeping with growing demand for fun-to-drive, in-wheel motor system (IWMs) of this e-4WD vehicle varying torques to each wheel independently is mainly aimed at improving the cornering performance on the high tire-road friction surface. The main purpose of this dissertation is to develop an implementable lateral motion control considering practical issues of actual e-4WD vehicle application, such as price competitiveness, driver’s ride comfort, light computational burden, high availability, and reliable control accuracy. In particular, the drawback of the simple feedback control in IWM system is that high-gain feedback may cause a large oscillatory response of IWM torque, which worsens the driver’s ride comfort, so that this dissertation proposes model-based control with a feed-forward control term.
For development of the model-based control for vehicle lateral motion, this dissertation introduces new methods for vehicle state estimation utilizing the competitively priced sensor fusion using in-vehicle sensors and low-cost standalone GPS. The main estimation targets are generally unmeasurable vehicle states, i.e. vehicle sideslip angle, tire cornering stiffness, vehicle heading angle, and precise vehicle position. To estimate the sideslip angle, an interacting multiple model (IMM) Kalman filter is developed, which includes a sensor offset compensator and extended Kalman filters (EKFs) based on multiple vehicle models. To properly combine the outputs of these model-based EKFs, a weighted probability of each model based on the stochastic process is designed, which reflects the characteristics of the vehicle models in real-time. Also, the observability of this IMM Kalman filter is checked by observability functions of nonlinear systems. As well as the sideslip angle, heading angle and tire cornering stiffness values of the front and rear axles are simultaneously estimated. Utilizing the estimation results of this IMM Kalman filter, a new adaptive Kalman filter with rule-based logic provides robust and highly accurate estimation of the vehicle position. It adjusts the noise covariance matrices in order to adapt to various environments, such as ever-changing GPS conditions. The performance of the proposed vehicle state estimation in various driving scenarios is verified using a test vehicle, and its superiority is confirmed through a comparative study.
Thus, the estimated vehicle state variables are transmitted to the vehicle lateral motion control. Two independent schemes of lateral motion control with different control targets are designed separately with the common goal of enhancing the cornering performance of the e-4WD vehicle. The control targets of schemes 1 and 2 are a yaw rate reference and a desired path profile, respectively.
The first scheme of the lateral motion controller for yaw rate tracking is aimed at neutral-steering which leads to the improvement of vehicle cornering agility. A smooth sliding mode controller (SMC) based on the vehicle bicycle model is utilized for yaw rate tracking. This smooth SMC with the feed-forward control term actively reflecting the yaw rate reference improves both convergence rate of control action and vehicle cornering agility.
The second scheme of the lateral motion controller for path tracking can assist the vehicle in following the desired path profile. Both a lateral distance error and a heading angle error between the desired path profile and the vehicle are controlled at the same time. A model predictive control (MPC) is selected as the controller, which derives optimal control input considering both state and input constraints in the e-4WD vehicle. Due to the advantage of this MPC to predict the vehicle's future dynamic behavior a few seconds ahead, it is possible to output a more preemptive and stable control input for vehicle path tracking.
Finally, a novel Daisy-chaining allocation, which is suitable for redundant actuator configuration, is utilized to distribute the desired yaw moment to front left and right IWMs of the e-4WD vehicle. It elaborately reflects both the characteristics of IWM and the tire friction circle. Noteworthy, it is a practical and intuitive method aimed at real-car application.
Using the CarSim simulation, the effectiveness of the proposed lateral motion controllers is verified. Thereafter, in real-car based experiments with various driving scenarios, it is confirmed that some evaluation factors in terms of the cornering performance are improved by comparing with conventional control algorithms. The following main contributions make this dissertation be a meaningful solution to enhance the cornering performance of e-4WD vehicle: 1) use of competitively priced sensor fusion 2) improvement of both smoothness and control accuracy due to model-based control with vehicle state estimation, 3) consideration of practical issues for real-car application.