Developer Notes
"WiThRay is currently under development to support Google Maps."
Most of the work is done in Manhattan.
WiThRay was based on the hexahedron-based 3D modeling data.
WiThRay is now targeting to conduct on 3D map data imported from Google Earth.
Working Environment: Intel(R) Core(TM) i7-10700F / NVIDIA GeForce RTX 2060 / 32GB RAM / MATLAB R2023b / Blender 3.6
2024. 1.24.
Try to import Google Earth map data to MATLAB (using Blender Blosm).
2024. 2. 6.
Success to import Google Earth map data to MATLAB. (Thanks to Dev Creates providing a tutorial [link])
[Map data imported to Blender]
[Map data imported to MATLAB] MATLAB does not support the UV texture mapping.
2024. 2.22.
Try to conduct the semantic segmentation (K-mean clustering).
2024. 3. 7.
Apply 3D blob for segmentation. (3D blob function provided by MATLAB is possible on voxels.)
Convert the polygon image to the voxelized iamge.
First, attempt to process the 3D blob, and then proceed with the classification.
Later, the procedure has been revised to first perform semantic segmentation, followed by the 3D blob processing (revised on March 21, 2024).
Create Manhattan 3D map data (hexahedron-based modeling) for WiThRay.
Image on the right side is the WiThRay ray-tracing result for the channel paths possible having two reflections and two diffractions.
This scenario will continue to be considered in future development.
2024. 3.21.
Use MATLAB Lidar toolbox for semantic segmentation (MATLAB supports PointNet++, which is a classification method for cloud points.)
Labeling the point cloud
Buildings, vegetation, and ground
Try to learn PointNet++
It does not work well.
It might be because of the lack of training data.
3D blob for the point cloud (vertices) that is classified as buildings.
The 3D blobs obtained here serve as higher-level objects of the point cloud (vertices), utilized to the hierarchical ray-tracing algorithm.
Implement one-point reflection.
It needs Blockage Determination Algorithm.
Possibly consider MeshNet-SP.
At the best of my knowledge, MeshNet-SP seems to be one of the best approaches for semantic segmentation on 3D city maps.
2024. 4. 3.
Develop hierarchical blockage determination algorithm.
2024. 4.17.
Implement UV mapping code in MATLAB.
Implement the ray-tracing algorithm for multiple reflection (Still, the maximum number of reflections is 2.)
Issue) Number of cases to explore for multiple reflections: (number of faces) ^ (number of reflections) → "Exponentially grow!"
Sol. 1) Reduce the number of cases to explore
Sol. 2) Parallelize the operation using tensor structure.
First and last reflecting surfaces should be connected to Tx or Rx nodes directly.
Channel includes...
Direct (line-of-sight) path
Indirect (non-line-of-sight) paths
One-point reflection path
Two-points reflection path
Yellow-colored lines are under the blockage.