Real-Time Drift-Driving Control for an Autonomous Vehicle: Learning from Nonlinear Model Predictive Control via a Deep Neural Network

A drift-driving maneuver is a control technique used by an expert driver to vehicle along sharply curved path or slippery road. This study develops nonlinear model predictive (NMPC) method for the autonomous perform drift and generate datasets necessary training deep neural network(DNN)-based controller. In general, NMPC based on numerical optimization which difficult run in real-time. By replacing previously designed with proposed DNN-based controller, we avoid need complex of control, thereby reducing computational load. The performance developed data-driven controller verified through realistic simulations that included scenarios. Based results simulations, showed similar tracking original controller; moreover, can demonstrate stable computation time, very important safety critical objective such as maneuver.