feat: create JuMPControlProblem for optimal control #3549
Conversation
|
@ChrisRackauckas assuming this passes stuff I think it's ready. |
|
Requires SciML/SciMLBase.jl#996 |
| @testset "ODE Solution, no cost" begin | ||
| # Test solving without anything attached. | ||
| @parameters α=1.5 β=1.0 γ=3.0 δ=1.0 | ||
| @variables x(..) y(..) |
There was a problem hiding this comment.
Do I have to define the variables to optimize over using x(..)? What if I want to solve an OC problem with components defined in MTKstdlib? Those components have x(t) variables. It would be nice to have a test for that use case as well since component-based modeling would be the fundamental reason to use MTK over any other tool
There was a problem hiding this comment.
No, that's just if you also want to evaluate them at a point, since this just late binds t. We could also make an evaluate operator or something that is the same as evaluate(x(t), 0.2) == x(0.2)
| function is_bangbang(input_sol, lbounds, ubounds, rtol = 1e-4) | ||
| bangbang = true | ||
| for v in 1:(length(input_sol.u[1]) - 1) | ||
| all(i -> ≈(i[v], bounds[v]; rtol) || ≈(i[v], bounds[u]; rtol), input_sol.u) || | ||
| (bangbang = false) | ||
| end | ||
| bangbang | ||
| end |
There was a problem hiding this comment.
| function is_bangbang(input_sol, lbounds, ubounds, rtol = 1e-4) | |
| bangbang = true | |
| for v in 1:(length(input_sol.u[1]) - 1) | |
| all(i -> ≈(i[v], bounds[v]; rtol) || ≈(i[v], bounds[u]; rtol), input_sol.u) || | |
| (bangbang = false) | |
| end | |
| bangbang | |
| end | |
| function is_bangbang(input_sol, lbounds, ubounds, rtol = 1e-4) | |
| for v in 1:(length(input_sol.u[1]) - 1) | |
| all(i -> ≈(i[v], bounds[v]; rtol) || ≈(i[v], bounds[u]; rtol), input_sol.u) || | |
| return false | |
| end | |
| true | |
| end |
| JuMPControlProblem(sys::ODESystem, u0, tspan, p; dt) | ||
| Convert an ODESystem representing an optimal control system into a JuMP model | ||
| for solving using optimization. Must provide `dt` for determining the length |
There was a problem hiding this comment.
dt is presumably not defined as "determining the length of the interpolation arrays". Instead of mentioning why a variable is required, maybe mention what the variable represents?
will add optimal control tests here when the interface for inputs and stuff is settled.