Inside GNSS Media & Research

JUL-AUG 2019

Issue link: https://insidegnss.epubxp.com/i/1148308

Contents of this Issue

Navigation

Page 52 of 67

www.insidegnss.com J U L Y / A U G U S T 2 0 1 9 Inside GNSS 53 the result using WLS method with the mean positioning error of 10.23 meters. However, the positioning error can go up to more than 30 meters in some epochs which can be seen in Figure 12. After applying the proposed perception aided NLOS correction method, the mean posi- tioning error decreases to 7.81 meters. Moreover, all the positioning errors are less than 30 meters. e trajectory during the test is shown in Figure 13. In the le- hand and right-hand side of the trajecto- ries, the degree of urbanization is shown using corresponding skymask which pres- ents only very limited sky visibility. As the NLOS correction is not available all through the test, the Figure 14 shows the trajectories only when the NLOS cor- rection is applied. We can find that major- ity of the NLOS corrections occurred when the experimental vehicle drives past the dense street shown in Figure 13. Table II shows the GNSS positioning results. We can find that the mean GNSS posi- tioning error decreasing to 7.13 meters. Moreover, the standard deviation also decreases slightly with the assist of the proposed method. Interestingly, we can find that the mean positioning error using the WLS method increases slightly from 10.23 meters in Table I to 11.01 meters in Table II. is means that the applied weighting scheme performs worse in more dense urbanized scenarios (the two dense streets are shown in Figure 13). e improved GNSS positioning results shows the effectiveness of the proposed percep- tion aided GNSS positioning method. e proposed method relies on the result of the object detection (building detection in this paper).With the increasing perception accuracy over time, more and more envi- ronmental information can be perceived with the on-board sensors of autonomous driving vehicles. e proposed method can perform better in improving the per- formance of GNSS positioning. Conclusions, Future Work In this article, the authors demonstrate the use of LiDAR perception to aid GNSS SPP. First we detect the building using the LiDAR point cloud data and extend its height according to the height list. en, FIGURE 10 Building detection result using the proposed 3-D point clouds segmentation method. The 3-D bounding boxes represents detected buildings. The 2-D black boxes denote the surrounding dynamic objects. The 2-D purple box represent the ego-vehicle. FIGURE 11 Illustration of a failure of building detection due to the excessive dynamic objects. The 3-D bounding boxes represents detected buildings. The 2-D black boxes denote the surrounding dynamic objects. The 2-D purple box represent the ego vehicle. FIGURE 12 Positioning error of the GNSS before and after adding the NLOS correction, and simply using least square in the tested urban canyon. The red curve indicates the positioning error using WLS, blue curve denotes the positioning based on proposed NLOS correction. The green curve shows the result using LS.

Articles in this issue

Links on this page

view archives of Inside GNSS Media & Research - JUL-AUG 2019