Open schillic opened 3 years ago
The minimally disrupting fix might be to pull Δt = (zero(N) .. δ) + Δt0
out of the reach_foo!
functions and pass it as an argument; if the algorithm works in discrete time, we have to pass Δt = zero(N) + Δt0
. We can decide based on a new is_dense_time
method applicable to <:AbstractContinuousPost
.
It would be better to only store the time points (which may require to add a new flowpipe type).
We can instantiate reach-sets with time points as well. It will still be wrapped as a (degenerate) interval, but that shouldn't matter.
Currently with
approx_model=NoBloating()
one receives a flowpipe with time intervals for each reach set. Thus when plotting such a flowpipe in time, it looks like a continuous-time result. It would be better to only store the time points (which may require to add a new flowpipe type).