Abstract
The paper presents a racing car setup evaluation support system that uses nonlinear model predictive control to emulate driver control actions. Simulation plays a crucial role in car competitions for reducing risks and costs of real track tests. Racing teams usually have their own simulation environments and the related technical literature is not of public domain. Commercial solutions are available with the related technical literature.
The proposed system is composed of: a simulator implementing vehicle model and control system (driver actions), a neuro-fuzzy subsystem estimating the vehicle performance changes resulting from setup modifications and a database storing simulations results (figure 1).
The implemented vehicle model is very detailed and includes 10 DOF (longitudinal and lateral vehicle translation, sprung mass vertical translation, yaw, roll, pitch, axles vertical translation and roll), which describes accurately the nonlinear behavior of a four wheels vehicle with elastic suspensions. This model includes: the semi-empirical nonlinear model defined by Pacejka equations, that describes tires behavior in terms of lateral and longitudinal forces with combined effect, powertrain dynamics, including gear changing and clutch, brakes dynamics, aerodynamic and friction effects.
Vehicle control system, which emulates the driver, is based on a nonlinear model predictive control (NMPC). The chosen control strategy includes main vehicle model nonlinearities and fits real driver decision process. As a driver, the control system takes into consideration desired performance, knowledge of vehicle behavior and limits, vehicle state perception, and uses a receding horizon strategy. The control system involves forces generated by tires and their combined effects, vertical vehicle load distribution on tires, longitudinal load transfer due to traction and braking, lateral load transfer due to lateral acceleration. Receding horizon length is chosen considering required prediction accuracy and acceptable computational load. The control inputs, consisting in front steering angle, brake and throttle pedal positions, are obtained minimizing the vehicle deviation from the target path and the desired velocity. Overall driver performance is evaluated using indexes for path-following and velocity tracking; lower values characterize a better car setup.
Experimental tests showed a good correspondence between telemetry data and simulation results.
Keywords: Racing car simulation, Driver simulation, Car control system, Nonlinear Model Predictive Control