Coverage map of non-flat surface

[Jklein64]

Is it possible to generate a coverage map for a 2D surface that isn’t flat, such as a mesh terrain specified by a heightmap?

I’ve experimented with the Sionna coverage map function and have only been able to change the orientation of the plane on which the coverage map is computed, as opposed to computing the coverage at different heights. Below is an example of why this behavior is problematic for us:

Our current workaround idea is to place many (hundreds or thousands) of receivers scattered across the terrain mesh, and then use the path losses from the transmitter to each receiver to get the data we need. This will be slow, and it would be helpful to be able to evaluate the coverage map function at the varying heights instead. Is there something fundamental about the coverage map computation algorithm that makes this difficult?

[faycalaa]

Hi @Jklein64,

It is possible to set the position of the center of the coverage map using the cm_center parameter of Scene.coverage_map. However, it is currently not possible to compute coverage maps for non-planar surfaces. The existing implementation assumes that the measurement surface is flat when calculating the contribution to the coverage map of a path intersecting the measurement plane.

In principle, it would be possible to consider non-planar measurement surfaces, but we currently do not have plans to implement this feature.

You could consider computing multiple smaller coverage maps, each with different positions and orientations, to approximate the shape of the terrain.

[redcoatwright]

@Jklein64 have you managed to make any progress on the issue of using a non-planar coverage map geometry? If not, I'll be diving into the source code as well to try to get it working.

[Jklein64]

@redcoatwright our initial approach had some pretty hefty drawbacks (such as requiring edits to most of the coverage map solver), but we recently transitioned to another approach based on textures that seems much more promising. Here are some images from recent experiments:

I plan to integrate this into Sionna and the coverage map solver soon. This method should require much less codebase updates than the previous approach, so fingers crossed it all falls into place nicely 🤞

[redcoatwright]

@Jklein64 is there any chance I can check out your fork where you're working on this? Trying to see if I can make this work for my use-case

[Jklein64]

@redcoatwright I realized I need the surface area of each cell as part of the coverage map computation, which is hard to compute for non-flat measurement surfaces. I've designed my own algorithm for this and am currently testing it, but it'll be a bit before everything is integrated together. I'll let you know once I have a fork you can look at!

[jhoydis]

Hi all,

Please note that the implementation of coverage maps has significantly changed in the new Sionna 1.0. Non-flat coverage maps are still not supported, but it is definitely an interesting feature we should consider for a future release.

[Jklein64]

Support for arbitrary mesh measurement surfaces in Radio Maps has been added in Sionna 1.1.0. Its implementation uses each mesh face as a “cell” instead of generalizing the grid of cells method (from planar Radio Maps) to work with non-flat measurement surfaces. This works, but the results are not exposed as a grid, and it's difficult to adjust the resolution of cells (since there is no analog to the cell_size parameter).

These two features, (1) a 2D path gain tensor and (2) a cell_size parameter, were very important to us. To handle this, I created another Radio Map solver that takes Digital Elevation Maps (DEMs, aka heightmaps) as input and evaluates the path loss on a "grid" draped over the heightmap. See this PR for more info.

Hope this helps! @redcoatwright @Mgomez-01