Closed schillic closed 6 years ago
This is what happens if we deactivate clustering for the second round. It does not look smaller than the first round, but also at least not bigger.
Here are two sample successor reach tubes (clustering deactivated).
Filtered oscillator, ApproximatingDiscretePost
:
plot(sol) # whole flowpipe
plot!(Reachability.ReachSolution(sol.Xk[1:137], Options(:plot_vars=>[1,2])), opacity=0.5) # first reach tube
plot!(Reachability.ReachSolution(sol.Xk[348:483], Options(:plot_vars=>[1,2])), opacity=0.5) # fifth reach tube
plot!(Reachability.ReachSolution(sol.Xk[138:138], Options(:plot_vars=>[1,2])), opacity=0.5) # successor of first reach tube
plot!(Reachability.ReachSolution(sol.Xk[484:484], Options(:plot_vars=>[1,2])), opacity=0.5) # successor of fifth reach tube
As we can see, the successor of the smaller set is bigger than the successor of the bigger set.
Also, notice that the sets are shifted to the top. I expected that the rectangles contain the end of the reach tube because here we do not have any assignments.
Example: filtered oscillator Even though the jump starting from the purple reach tube should land in a subset of the green reach tube (which was the jump target from the orange reach tube), we obtain the blue reach tube.