Closed 0Yassine0 closed 1 week ago
@0Yassine0 When you have finished to add the problems, you can make a PR.
I would suggest we add the full list, including the problems where OC is not converging.
As a reference, here is the list of all the problems we have so far:
Yess I only used the models where we have the same results (nvar, ncon, objective value..) for the same nh. I'll add the full list to the package if you're okay with that. For the tests I checked that the solver gives an optimal solution, should I add a list to exclude the other models or I need to change the test?
Let's add the full list.
For the tests, we should check that:
MOI.LOCALLY_SOLVED
for JuMP, the equivalent for OC/NLPModels. @test_broken
if Ipopt is not converging on a particular instance. That way, we can track which problems are failing. @PierreMartinon do we have an equivalent of MOI.LOCALLY_SOLVED
?
@PierreMartinon do we have an equivalent of
MOI.LOCALLY_SOLVED
?
Solving with OC has this field in its return value: sol.message
, so we can consider "Solve_Succeeded" as an equivalent of MOI.LOCALLY_SOLVED
?
@0Yassine0 not sure. There is also sol.success
but I am not sure this field is used.
From COTS.jl package, we will add the problems that works perfectly with
JuMP
andOptimalControl
: The Hanging Chain, The Hang Glider, The Robot Arm, The Goddard Rocket, The Particle Steering, The Cart Pendulum, The Moonlander, The Dielectrophoretic Particle, The Ducted Fan, The Double Oscillator and The Electrical Vehicle.