search

I24-MOTION / VT_tools

Toolbox for creating spatiotemporal speed fields and generating virtual trajectories within these fields. Virtual trajectories are useful for analytics because they are smooth, complete, and can be generated at specific and uniform times. This code accompanies an article presented at IEEE-FISTS 2024 (Ji 2024).
https://i24motion.org
Other
3 stars 2 forks source link

The reason for multiplying 2 at the ASM.py beta_free_flow and beta_cong_flow and setting hyperparatmeters. #3

Open seungwooham opened 4 months ago

seungwooham commented 4 months ago

Hi, first of all, thanks for sharing the rich material for I24 data processing.

While I was checking out the code, I found that the return value was multiplied by 2 at the beta_free_flow and beta_cong_flow functions.

I guess the authors of this repo would have some specific reason for this, because in the original paper (An adaptive smoothing method for traffic state identification from incomplete information), the multiplied coefficient 2 is omitted.

image

In the code, we can find out the return value as

return np.exp(-(2np.abs(dx)/x_win + 2np.abs(dt)/t_win)) return np.exp(-(2np.abs(dx)/x_win + 2np.abs(dt)/t_win))

Could you please explain the reason for this?

In addition, it would be really great if you could add some reason for setting the hyperparameters.

Such as smooth_x_window $\sigma$, smooth_t_window $\tau$, c_free, c_cong, and V_c.

Thank you!

junyi9 commented 3 months ago

Hi,

Thank you for reporting the issue. It appears to be a bug in the code, and I will address it in the next release. ASM.py will be updated. Regarding the hyperparameters, I did not follow the original paper's approach, where the observation data point is with a larger interval in that paper. For the next release, I will tune the hyperparameters accordingly.

Thanks again for your valuable feedback!