VEHICLE DYNAMICS MODELING FOR AUTONOMOUS DRIFTING AND CLOTHOID BASED WAYPOINT INTERPOLATION

The advent of autonomous vehicles necessitates a redefinition of road safety regulations, considering a controller can possess better driving skills than an average person. The work presented here partly focuses on a vehicle dynamics model development to help imitate and control vehicle drifting maneuvers. As we see, a professional driver drifting through the traffic while keeping the car safe, it can be utilized to avoid accidents at high speeds, if required. Although drifting can produce higher yaw rates than the regular driving regime, these control capabilities have not yet been exploited in the current automotive control systems. Therefore, this report focuses on developing a vehicle dynamics model to simulate the drifting of an autonomous vehicle that utilizes this high yaw rate property. Drifting control capabilities will increase the vehicle\u27s ability to avoid collisions scenarios where higher than typical yaw rates are required. The other part of the report uses vehicle kinematics equations to generate feasible path planning algorithms for any autonomous vehicle. If we constraint these vehicle kinematics equations for linearly varying curvature of the path, they are called Clothoid curves. This linearly varying curvature is analogous to having a continuous lateral acceleration on the vehicle while cornering, i.e., avoiding jerks. Here, Clothoids are used for the interpolation of waypoints along the path. A mission planner defines waypoints a few meters apart from each other; then, GPS coordinates are used to check whether the vehicle is following these waypoints correctly. Therefore, waypoint interpolation is required for continuous feed of track coordinates to the controller to make faster corrections for the cross-track error and not to wait every time for a low rate GPS signal. Also, the Clothoid curves are used in road construction, thus also provides the ability to create a road model along with a vehicle model which can further be used for better state estimation