Time optimal control strategy
The aim of this project is to further develop the CARS system constructed by a student group during the summer of 2014. The main goal of the project is to decrease lap time, while maintaining stable behavior of the cars. A time-optimal car trajectory is calculated, using optimal control principles, that is used as reference trajectory for the controller. Further improvements to the speed reference are made through an adaptive, learning algorithm to account for changes/uncertainties in car parameters and track conditions. Furthermore, the project includes adding functionality for two autonomous cars, a real-time implementation of a particle filter for state estimation and synthesis of a theoretical car model.
The time-optimized CARS system now achieves a best lap time of 4.45s on ideal track conditions, to be compared with the record of the former system of 5.9s. With the new adaptive speed reference the cars achieve a best lap time of 5.03s. There has been hardware and software additions to allow for two autonomous cars to drive simultaneously. A theoretical model of the car was parameterized and validated and a simplified version was implemented in the state estimation filters. The particle filter works well for state estimations with similar performance in terms of tracking as the extended Kalman filter but is computationally more expensive.