Experimental Assessment of GNSS-based Vehicle Positioning Accuracy using 3-D SLAM Reference

Many vehicular and ITS applications require a highly accurate real-time vehicle position estimate for ensuring their safety and efficiency. The current limitation of vehicle positioning research is the lack of reliable position reference regardless of the surrounding environments. To mitigate this problem, we use the position estimate of three-dimensional simultaneous localization and mapping (3-D SLAM) as the ground truth. Then, we experimentally assess the positioning accuracy of three GPS receivers, i.e. the normal GPS, SBAS-GPS, and RTK-GPS, in two different geographical environments. From the numerical results, the RTK-GPS is shown to achieve the best positioning accuracy (with RMSE of 1.18 m) in the PNU campus with a good sky view, but surprisingly, its positioning accuracy in a high-rise apartment complex is the worst (with RMSE of 42.6 m) presumably due to the blocking and multi-path reflection of the GPS signals.