Safety margin issue

[shengjiex98]

This is not a real PR, but a minimal working example of #26. Running

julia --project test/safety_margin_issue.jl

multiple times yields different results each time.

I realize that this is after a bunch of changes that I've made to the schedule synthesis part, so it could be something that I did there broke it. Still, would be very helpful to find out the cause of this!

[Ratfink]

It would be nice if we could get a more minimal example that simply shows the problem on the main branch to verify that it isn't because of all the other changes here. Shouldn't be any need to make a whole branch for it, just a short script I could copy and paste should suffice. I could take a shot at it myself if you don't have the time, just let me know. 🙂

[shengjiex98]

I'll take a look today and get back to you. I think last time I tried it was only showing up in my branch so maybe something there messed things up. I'll follow up after I try it out on main.

[shengjiex98]

Seems like I cannot reproduce the problem in main and the issue is likely in the changes I made after. From my initial investigation, the issue seems to stem from bounded_runs_iter() sometimes exits earlier even when using the same constraint. I.e., running

reachable = bounded_runs_iter(a, z_0, n, H, safety_margin=d_max)
if size(reachable, 1) != H+1
    @info "Size mismatch"
    return d_max
end

10 times in a row sometimes enters the if scenario even when none of the parameters have changed. My next step is to do a diff between the two branches on bounded_runs_iter and see if I made any changes there