Inside GNSS Media & Research

JUL-AUG 2019

Issue link:

Contents of this Issue


Page 49 of 67

50 Inside GNSS J U L Y / A U G U S T 2 0 1 9 matching faces other challenges as mentioned in (Groves 2012, Additional Resources). e recent state-of-the-art range-based 3DMA GNSS can correct most of the pseudorange measurement affected by NLOS receptions (see again Hsu 2018 in Additional Resources). However, the computational load of the ray-tracing simulation is immense as the simulations are required in each hypothesized position. In addition, an accurate prior-known receiver posi- tion is required by the 3DMA GNSS. To address these two issues, as an example, we present a novel method to detect the GNSS signal blockage caused by surrounding buildings and correct the NLOS pseudorange measurements based on the perceived environment fea- tures by the sensors installed on autono- mous driving vehicles. Methodology To estimate the geometry and pose of the buildings relative to GNSS receiv- er, a surface segmentation method is employed to detect the surrounding building walls using LiDAR 3-D point clouds. The building boundaries are extracted and extended by the building height in a skyplot to identify the NLOS affected ones from all the measure- ments. Innovatively, the NLOS delay in pseudorange can be modelled and cor- rected. Weighted least square (WLS) is used to cooperate the corrected NLOS and healthy pseudorange measure- ments. Figure 5 shows the flowchart of the proposed method. e steps of the proposed method are as follows: STEP I: Building Surface Identification and Extension To detect the top edges of buildings (TEBs) and obtain the corresponding distances between the GNSS receiver and buildings, a point cloud segmentation method is employed to implement the building sur- face detection. To distinguish the building surface from the unordered points set and determine the distance from GNSS receiv- er to the building surface, two steps are needed: the segmentation and building surface identification. e segmentation and surface identification are described in detail as shown in Algorithms 1 and 2 in the paper by Wen, et alia in Additional Resources, respectively. The output of Algorithm is the points clusters shown in the le-hand side of Figure 6 and we do not know which cluster belongs to the build- ings class. e segmentation in Algorithm 1 clusters the points into bounding box U i which can be described as following: (1) where , and denote the position of the bounding box in x, y, and z directions in LiDAR coordinate system, respectively; roll , pitch and yaw denote the orientation of bounding box in LiDAR coordinate system. is the length, is the width and is the height of the bounding box. To effectively identify the bounding box representing the building surface which can result in GNSS signal reflec- tion and subsequent NLOS receptions, building surface identification method is needed. By the Algorithm 2, the build- ing surface can be identified shown in the middle of Figure 6. e height of the bounding box representing building sur- face can be extended to the real one. In fact, this building height extension can be omitted if a sky-pointing camera is used (Suzuki and Kubo, Additional Resources). e bounding box is extended from rect- angle ABCD to rectangle CDEF as can be seen in the right-hand side of Figure 6. en, the boundary parameters for the bounding box corresponding to build- ing surface are denoted by line segment denoted as , the matrix of the boundary. To represent the building, two points, E and F, are required. e is structured as follows: (2) In this case, the TEBs of the buildings represented by the . STEP II: Projecting the Top Edges of Buildings into a GNSS Skyplot To detect NLOS, visibility of satellite needs to be determined based on the extended TEBs ( ). The relative position of the GNSS receiver to satel- lites and to building surfaces needs to be transformed into the same representa- tion, the Skyplot. Satellite position can be easily indicated in the Skyplot represen- tation based on corresponding elevation and azimuth angles. A transformation matrix should be employed for building surface boundaries transformation from 3 dimensions coordinate to 2 dimensions coordinate. e transformation is con- ducted as the following formula. (3) FIGURE 6 Illustration of point sets segmentation and building surface identification. Blue box ABCD represents the initially detected building surface. Blue box CDEF represents the extended building surface. FIGURE 5 Flowchart of the proposed method of GNSS SPP with NLOS correction. SINGLE POINT POSITIONING

Articles in this issue

Links on this page

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