Measurement quality control aided multi-sensor system for improved vehicle navigation in urban areas

Integration of the global positioning system (GPS) with other sensors to improve positioning and navigation is a viable solution for intelligent vehicles. In particular, integrating GPS with an inertial measurement unit (IMU) and vision sensor is a common consideration for enhancing the performance of vehicle navigation systems. However, the accuracy and availability of GPS are seriously degraded by Non-line-of-sight (NLOS) reception and multipath interference in urban environments. Combining the unreliable GPS positioning results with other sensors will degrade the performance of the integrated navigation system. Therefore, we propose a novel measurement quality control algorithm for a loosely coupled GPS, IMU, and stereo visual odometer (VO) integration for urban vehicle navigation. In the proposed algorithm, an adaptive fusion strategy based on geometric dilution of precision (GDOP) and a measurement fault detection and exclusion (FDE) based on the K-means clustering considering predicted pseudorange errors are employed to mitigate NLOS/multipath effects. The validation of the proposed integrated solution incorporating the measurement quality control is carried out. The results show that the algorithm achieves positioning accuracy in terms of 3D root mean square error (RMSE) of 2.85 m and 7.17 m, in mid and urban canyons, respectively. This level of performance is superior to that of the traditional GPS/IMU/VO integration. The corresponding improvements are 49.0% in the mid urban canyon, and 79.6% in the deep urban canyon respectively.