You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Say that you want to plot dimensions x3 and x4 and you do a decomposed analysis only for those two dimensions. Then the result will be a 2D flowpipe representing x3 and x4. However, the projection would try to access dimensions 3 and 4 (because those were the specified dimensions), which do not exist in a 2D flowpipe.
Current state
Internally we create inout_maps to map to the correct dimensions. But this is poorly integrated and at some places we just inject them.
Proposal
Let the ReachSet automatically handle these situations. They should receive the information which dimensions are stored and which of them are just taken as universal. (This is essentially what was proposed in #275.) Essentially this internal information is equivalent to the inout_map, but we can remove some hacks.
The text was updated successfully, but these errors were encountered:
We add a new type for low-dimensional ReachSets (suggested name: SparseReachSet) using type composition.
We add a project function with methods for ReachSet and SparseReachSet which, given a list of dimensions, returns a set projected to those dimensions. Dimension 0 for time is possible.
Example
Say that you want to plot dimensions
x3
andx4
and you do a decomposed analysis only for those two dimensions. Then the result will be a 2D flowpipe representingx3
andx4
. However, the projection would try to access dimensions 3 and 4 (because those were the specified dimensions), which do not exist in a 2D flowpipe.Current state
Internally we create
inout_map
s to map to the correct dimensions. But this is poorly integrated and at some places we just inject them.Proposal
Let the
ReachSet
automatically handle these situations. They should receive the information which dimensions are stored and which of them are just taken as universal. (This is essentially what was proposed in #275.) Essentially this internal information is equivalent to theinout_map
, but we can remove some hacks.The text was updated successfully, but these errors were encountered: