Automated Tuning and Calibration of By-Wire Vehicles for Automated Driving Functions

The current process for setup and configuration of autonomous ground vehicles is a time consuming process which can often require a skilled engineer. This process includes measuring key sensor parameters and also tuning control parameters so the vehicle will operate properly. This proposed research will develop techniques for self-calibration of sensor parameters and automatic tuning of control system parameters for autonomous ground vehicles. The sensor parameters will include sensor locations, orientations and scale factors required for accurate measurement use. An Extended Kalman filter framework is proposed that allows key sensor states to be augmented into localization filter that exists on the vehicle. The automatic tuning will improve the control of the vehicle while driving in autonomous mode. A Model Predictive Control framework is proposed to make this possible. The MPC gains will be improved by monitoring the actual trajectory of the vehicle against a benchmark optimal trajectory. The difference between the paths is used as feedback into the controller gains calculation. The actual gains used in the MPC will converge to optimal values over time. This work will be implemented in simulation and also initial results will be obtained from actual vehicles.