Low-cost Lane-level Positioning in Urban Area Using Optimized Long Time Series GNSS and IMU Data

In this paper, we proposed a novel technique to realize accurate and robust position and pose estimation in a dense urban area. The technique make the best use of averaging effect to optimize long time (over several tens of seconds) series sensor data. Our proposed scheme uses just a low-cost GNSS receiver, a MEMS IMU, and a speed sensor. Evaluation tests in a Japanese urban area showed that our proposed scheme can realize robust lane-level absolute positioning results (2DRMS, 0.9 m). In addition, the standard deviation of the heading is 0.4°, and that of the pitch angle is 0.6°. Evaluation tests showed that the accuracy of our proposed scheme almost reached levels of the survey level mapping system, which is equipped with high-cost sensors. On the other hands, the total sensor cost for our prototype was only several hundreds of dollars. We believe that our proposed position and pose estimation scheme enables enhanced vehicle application to systems such as driver assistance systems, autonomous vehicle, and mapping systems.