Multi-scenarios multi-objective predictive controller for lateral control of autonomous vehicles

Path tracking is one of the key functions for autonomous driving. For a safe and comfortable ride, proper assistance should offer high accuracy and smooth tracking. To meet these requirements, Model Predictive Control (MPC) is proving to be a promising solution. However, classic predictive controllers may have limitations in practical challenges when requirements change due to diverse scenarios in a single ride. This work introduces a multi-model predictive control framework that integrates path tracking, operational safety (vehicle handling stability, environmental envelope), and passengers’ comfort (minimum lateral acceleration and jerk) and mediates among these sometimes-competing objectives by prioritizing safety. It can even temporarily reduce tracking quality without compromising safety if this is necessary to enhance passenger comfort. The optimal control problem is solved online using qpOASES and the controller is verified in the IPG CarMaker® high fidelity simulator using a highway exit scenario, demonstrating its ability to balance lane centering and passenger comfort while ensuring safety.