Vehicle Direct Yaw Moment based ESC and Wheel slip-based ABS Control Optimization using Data-Driven Approach and CADLM Lunar

Reduced-order modeling techniques (ROM) are applied to eliminate the components or modes that have little or almost no impact on the system's output. Thus creating a reduced-order system with nearly the exact behavior, but less complex and faster to undergo simulation or to operate in real-time. In this thesis, test vehicle data logs were analyzed to obtain insights, find optimal parameters, and match the Simulink vehicle model parameters to the actual test vehicle. Afterward, a longitudinal slip-based Anti-lock Brake System (ABS) is built and integrated with an Electronic Stability Control (ESC) through Direct Yaw Moment (DYM), which is realized via the Sliding Mode Control (SMC)technique. In order to create ROM models, datasets were collected from ABS and ESC simulations. They included control parameters as inputs and the desired simulation responses as outputs. Moreover, the datasets were passed to CADLM Lunar to find the best ROM technique for each output response. Similarly, Lunar was employed to optimize both the ABS and ESC control parameters. In the ABS, optimization goals were selected so that the wheels would maximize vehicle braking force and maintain steerability. On the other hand, the ESC optimization targets were to track the desired yaw rate and body slip angle as well as minimizing the speed loss, jerk effect, and lateral acceleration. Hence, laying down the conditions for vehicle skid prevention. The resulting output curves of the reduced model ABS were a close match of the simulation output with the same optimal control parameters. Furthermore, the ABS maintained the requested longitudinal slip within the required range during the braking period. Similarly, the ESC output responses of the reduced-order nearly matched the simulation outputs. Furthermore, when simulating large steering on the standard Sine and Dwell test without activating the ESC, the vehicle loses the necessary traction to perform the maneuver. On the other hand, when the ESC is active, it tracks the ESC suggested trajectory. Finally, the reduced model proved to be highly effective in terms of accuracy and time. In particular, the models produced by Lunar take practically no time to generate, while the same result takes several minutes to run in Simulink.