An Experimental Exploration of Low-Cost Sensor and Vehicle Model Solutions for Precision Ground Vehicle Navigation

Abstract

This thesis presents the use of a vehicle dynamic model (VDM) in conjunction with standard vehicle safety sensors to assist a low-cost Inertial Measurement Unit (IMU) and low-cost Global Positioning System (GPS) for fused precision ground vehicle navigation. Current production vehicles are outfitted with a multitude of sensors that are essential to safety packages such as: Electronic Stability Control (ESC), navigation, Adaptive Cruise Control (ACC), and lane tracking. The research presented in this thesis focuses on the sensors used with ESC and navigation, i.e. Wheel Speed Sensors (WSS), Steer Angle Sensors (SAS), IMU, and GPS. Each of the proposed measurement systems is plagued with detrimental characteristics that prevent standalone high precision and robust ground vehicle localization. The benefits of fusing GPS and an Inertial Navigation System (INS) are well documented, therefore a benchmark GPS/INS localization algorithm based on well known literature is created. The benchmark system is used as a comparison medium in order to directly quantify performance of the localization algorithm utilizing VDM measurements.

Initial research yields drastic improvements in performance of the GPS/INS/VDM localization algorithm over the benchmark GPS/INS algorithm. However, the GPS/INS/VDM algorithm demonstrates degraded performance in comparison to the GPS/INS algorithm during a few driving scenarios. A constant covariance for the VDM measurements in the EKF causes the localization algorithm to drift in moments of high dynamics. Further analysis of all measurement systems leads to the development of a Complementary Covariance Filter (CCF) for the VDM measurements that allows the localization algorithm to weight the VDM measurement update accordingly during periods of high dynamics. With the CCF improvements, the updated GPS/INS/VDM algorithm is able to achieve improved or at least equivalent performance over the baseline GPS/INS localization solution for all situations. Therefore, this thesis proves that the inclusion of the VDM system along with ESC sensors provides an optimal low-cost ground vehicle localization solution when considering production cost and performance.