Nonlinear Model Predictive Control for Optimal Motion Planning in Autonomous Race Cars

Abstract

Motorsport has historically driven automobile innovation by challenging the world's best car manufacturers to design, and develop vehicles that push limits of contemporary technology, and compete at physical vehicle limits. Autonomous driving is a rapidly evolving field that garners interest in industry, government, and research due to substantial improvements in road safety, and traffic flow. Autonomous racing is a byproduct of advancements in autonomous driving, and guarantees innovation in the field through the design of state-of-the-art perception, motion planning, and control algorithms developed to perform in fast-paced, multi-object environments at high speeds, operating at a vehicle's acceleration, and tire limits. We propose a high-level Nonlinear Model Predictive Control (NMPC) strategy incorporating a Pacejka tire model, and nonlinear vehicle dynamics in the global coordinate system with constraints based on track boundaries, and vehicle input limits for optimal motion planning to minimize lap time. The NMPC motion planner is evaluated in three real-world racetracks, Circuit of the Americas, Circuit de Spa-Francorchamps, and Autodromo Nazionale Monza for three race car classes, Formula 1 (F1), Le Mans Prototype (LMP1), and Grand Touring Endurance (GTE). The proposed NMPC strategy is shown to generate time-optimal trajectories for each vehicle class in the evaluated tracks, conforming to optimal racing lines demonstrated by professional racing drivers.