The GPS Vehicle Dynamics Laboratory focuses on the robust control of autonomous vehicles using GPS and Inertial Navigation System (INS) sensors. Our research has three main thrusts: sensor fusion/integration, on-line system identification, and adaptive control techniques and their application to vehicle dynamics and transportation. These methods can be used for such things as determining vehicle and driver models. Improved driver models could be used by a number of vehicle monitoring systems, i.e. safety systems that determine the effectiveness of a driver and increase road safety by removing fatigued or intoxicated drivers from the road. Vehicle modeling and state estimation is important in a number of current vehicle safety systems, such as ABS, traction control, and stability control. Additionally, future vehicle safety systems, such as driver assisted systems, adaptive cruise control, and even full autonomous lane-keeping, require precise vehicle models.

The first part of our research is to investigate methods for better calibration of the INS errors while GPS measurements are available. This will improve performance of the INS unit during periods when the GPS signal is obstructed (as well as between GPS measurements). Improved performance will be sought by including dynamic models of the vehicle system and incorporating these dynamic constraints with low-level INS/GPS measurements. Carrier-phase GPS signals, in conjunction with the system model, will be used to accurately calibrate the INS model and its errors. This precise calibration will provide a dead reckoning system, initialized using GPS, capable of providing accurate estimates of the vehicle states (position and attitude) for the continuous control of the vehicle during GPS outages. The integration of INS and GPS can be used to provide an unbiased, high-update estimate of vehicle states such as position, velocity, and attitude. This blended solution thereby provides accurate data for modeling autonomous vehicles. The ability to accurately determine the vehicle states as well as the vehicle model on-line, during changing environments, will in turn lead to an increase in the control performance of a vehicle.

Finally, our research focuses on adaptive control and estimation algorithms for autonomous vehicles. On-line system identification techniques capture the changing parameters of the systems, which can be used to adapt the control and estimation algorithms. Once techniques for using the GPS/INS solution to perform on-line identification have been developed, methods that adapt, or self-tune, optimal controllers and estimators (such as LQR and Kalman filters) can be investigated. The adaptation of the control and estimation algorithms to the continually identified model parameters will lead to accurate and robust performance of these autonomous systems.