Autonomous Cornering at the Limits: Maximizing a “g-g” Diagram by Using Feedforward Trail-Braking and Throttle-on-Exit

This paper explains the design of a feedforward longitudinal controller that will drive an autonomous vehicle through a corner at the limits of tire adhesion. The feedforward longitudinal algorithm applies appropriate amounts of braking force and throttle during cornering to mimic a racecar driver trail-braking into a corner and applying the throttle during corner exit. A “g-g” diagram combined with a clothoid map is used to estimate the feedforward longitudinal command. By introducing this feedforward algorithm into the controller, significant improvements in lap time and cornering exit speed were observed. Analyzing the controller performance during cornering also provides an insight to the future development of the feedback longitudinal controller. The knowledge gained from designing an algorithm that controls a vehicle at the limits can be applied to future driver assistance systems to make cars safer, yet still fun to drive.