Development, Implementation and Testing of a Yaw Control system for low friction applications

This project aims to design and develop a vehicle yaw rate control for low-grip applications. A detailed review of the state of the art is carried out, leading to the choice of developing a PI controller and an advanced Non-Linear Model Predictive Control (NMPC). The NMPCs were implemented using the ACADO Toolkit, a C++ software environment and algorithm for automatic control and dynamic optimisation. The NMPC is a model-based control. Exploiting the vehicle model and the implementation of the Pacejka Magic Formula (MF) for tire force generation, it adapts properly to varying road conditions (dry, wet and snow-covered), through the awareness of the road friction coefficient. Three NMPCs were developed with increasing vehicle model complexity: a simple bicycle model, a twin-track model, and an advanced twin-track model with torque-split logic, which incorporates the yaw torque actuation method in the predictive controller's logic. The limitations and possible improvements of each controller were investigated in the simulation environment of VI-CarRealTime, an advanced simulation software developed by VI-Grade. A vehicle model based on the specifications of an Alfa Romeo Giulia is designed in VI-CarRealTime. The powertrain of the VI-Grade reference model was modified by adding 4 independent electric machines, one for each tire. This configuration allows for complete torque-vectoring control, minimising the effort of the low-level controller dedicated to torque actuation. Results show that the NMPCs outperformed the PI controller in terms of stability and cornering performance under low-grip conditions. Moreover, the knowledge of road friction coefficient improved controller accuracy, as the controller's plant closely describes the real vehicle-road interaction, enhancing vehicle stability in high-speed cornering and increasing overall vehicle safety.