<b>VEHICLE DYNAMICS HANDLING ANALYSIS FOR AUTONOMOUS RACECAR</b>

<p dir="ltr">The actual vehicle model is one of the most important aspects in autonomous cars. Understanding the vehicle becomes crucial for creating optimal control in the absence of a driver. High-fidelity vehicle models can be used to develop optimal steering controllers and gain insights into vehicle handling limits, which are essential for operating the vehicle at its performance limit during emergency maneuvers and planning efficient race lines in overtaking scenarios. By estimating these limits through machine learning-based parameter estimation, more confidence can be gained in high-speed overtaking maneuvers. Data-driven techniques are used to obtain a preliminary model using multiple regression (MR) for model fitting through data. A linear bicycle model serves as the structure for fit ting. The model is used to design a feedforward (FFW) controller for an Indy Autonomous Challenge racecar. By analyzing model fitting coefficients, vehicle parameters can be ex tracted and used to create a physics-based model that incorporates nonlinear tire behavior and aerodynamic downforce. Adding aerodynamics and nonlinear tire dynamics increases model fidelity to get accurate results for MR model fitting structure. Subsequently, an adaptive model is developed using machine learning. Deep learning techniques can help estimate key parameters like tire coefficients, their variation with speed, downforce, and the vehicles dynamic behavior, such as understeer characteristics and the influence of lateral dynamics on vehicle roll or bank angles.</p>