Vehicle localization using sensors data fusion via integration of covariance intersection and interval analysis

Achieving an innovative integrated sensor fusion architecture with a robust vehicle navigation and localization using an extended Kalman filter, interval analysis and covariance intersection that can overcome the uncertainty in the system model and sensor noise statistics. There are various approaches to the problem, but here the focus is on an approach which can guaranteed performance of sensor-based navigation. The guaranteed performance is quantified by explicit bounds of position estimate of a ground vehicle. Ground vehicles generally carry dead reckoning sensors such as wheel encoders and inertial sensors, to measure acceleration and angle rate, while obstacle detection and mapmaking is done with time-of-flight ultrasonic sensors. Most of these sensors give overlapping or complementary information and sometimes are redundant as well, which offers scope for exploiting data fusion. The purpose here is to achieve data fusion for ground vehicles with low-cost sensors by forming an intelligent sensor system. This is accomplished by combining the sensors' measurements and processing these measurements with data fusion algorithms. The algorithms are complementary in the sense that they compensate for each other's limitations, so that the resulting performance of the sensor system is better than its individual components.