Stone-Soup
Stone-Soup copied to clipboard
dstl
Any ideas on TO-MHT implementation?
Dear experts,
I am currently working on TO-MHT algorithms. I was wondering is it possible(convenient) to implement TO-MHT using current available modules?
I appreciate your valuable advice!
Sincerely, Chuan
We've got a lot of building blocks. I think we'd need to create a Tracker class, which doesn't have a DataAssociator but instead an Hypothesiser (which would include Gater classes), and then handles creation/pruning of track tree structure. We could then create different tracker classes, or maybe pass in different pruning classes, to create different variants.
It may make sense to have this track tree structure have a set like interface which is yielded from the tracker, which then yields the current time "best" tracks when iterating over. This way you could create a copy of the structure at each timestep if you wanted to understand how MHT evolved over time, as well as providing familiar interface to other components.
Thank you! I'll work on it after I discuss with my colleagues. Hope there will be a meaningful conversation going on.
Dear @Huang-Chuan , i am just curious to know if there is any update to this issue? Regards,
Dear @khurramkhalil, @Huang-Chuan and @sdhiscocks I am myself interested in such implementation.
I am having trouble using the already existing stone soup components to create the tracker as the initiator and deleter don't happen at the same time as when using GNN and JPDA associators.
Also wondering if any TreeTrack structure and MWIS solver have shown to be efficient.
Thank you, Albert
We are currently commissioning development of MHT into Stone Soup, so hopefully see something in coming months. I've reopened this issue so we can track it here.
@Bebert98 Maybe you could open a discussion with the issues you're having and we can try and help.
You may also be interested in #568. I've just been updating this to latest version and adding an example, which I'll push in draft form soon.