Differentiable Pathloss Parameter

[fanyimeng0]

Hi, I was wondering if you are interested in developing a differentiable path loss parameter. I am referring to the relationship between path loss and distance.

I have reviewed _compute_paths_coefficients and believe the current exponent is set to 2, but I think it might be meaningful to learn this parameter in certain environments.

Additionally, do you foresee any potential issues if I set this exponent as a tf.Variable? Thank you!

[jhoydis]

Hi @fanyimeng0,

There is no explicit pathloss parameter in Sionna RT. The fact that the power scales as 1/r^2 for LOS and specular paths is because we have spherical wave propagation. (You can check in the EM Primer that this is different for paths including diffraction, RIS, or diffuse reflections). This should not be modified and there would be also no physical justification for it as far as I know.

However, I imagine that you would like to have an additional distance-based pathloss, e.g., due to atmospheric absorption, rain, or something like this. One way to hack this into Sionna would be to transform the delay tau_i for every path into a distance (using the speed of light) and then add an additional pathloss based on this distance.