A Cascaded Optimization Approach for Modeling a Professional Driver's Driving Style

Abstract

In the context of minimum-time vehicle maneuvering, previous works have shown that different professional drivers drive differently while achieving nearly identical performance. In this paper, a cascaded optimization framework is presented for modeling individual driving styles of professional drivers. Therein, an inner loop model predictive controller (MPC) finds the optimal vehicle inputs that minimize a blended-cost function over each receding horizon. The outer loop of this framework is an optimization computation which finds the optimal weights for each local MPC horizon that best fit data obtained from onboard vehicle measurements of the targeted drivers to the simulation of the maneuver under the cascaded control. This cascaded optimization is exercised for a case study on Sebring International Raceway where two different professional drivers were able to achieve nearly identical lap times while adopting different driving styles. It will be shown that this framework is able to model key differences in style between the two drivers during a particular corner. The models of the individual drivers are then fixed, and another optimization is used to tune tire parameters to suit each driving style and illustrate the utility of the approach.