Your first problem
We take the clamped beam: a double integrator whose position must stay below a bound, with a quadratic cost. We build it with OptimalControl, solve it, and compare the objective to the reference of the collection.
using OptimalControl, OptimalControlProblems
using NLPModelsIpoptGet the problem
A problem is a function, called with the modeling as argument. With OptimalControl it returns a Problem: the continuous problem and its initial guess.
pb = beam(OptimalControlBackend())Problem beam (OptimalControl modeling, valid)
state dimension: 2
control dimension: 1
variable dimension: 0model(pb)Abstract definition:
t ∈ [t0, tf], time
x ∈ R², state
u ∈ R, control
x(t0) == [x₁_t0, x₂_t0]
x(tf) == [x₁_tf, x₂_tf]
x₁_l ≤ x₁(t) ≤ x₁_u
ẋ(t) == [x₂(t), u(t)]
∫(u(t) ^ 2) → min
The (autonomous) optimal control problem is of the form:
minimize J(x, u) = ∫ f⁰(x(t), u(t)) dt, over [0, 1]
subject to
ẋ(t) = f(x(t), u(t)), t in [0, 1] a.e.,
ϕ₋ ≤ ϕ(x(0), x(1)) ≤ ϕ₊,
x₋ ≤ x(t) ≤ x₊,
where x(t) ∈ R² and u(t) ∈ R.Solve it
The discretization is chosen in solve: here the trapezoidal rule on the default grid of the problem, an ADNLPModels model and Ipopt. The reference objectives of the collection are computed with the trapezoidal rule, which is not the default scheme of solve: give it.
N = grid_size_data(:beam, OptimalControlBackend())
sol = solve(
model(pb), :collocation, :adnlp, :ipopt;
init=initial_guess(pb),
grid_size=N,
scheme=:trapeze,
display=false,
)
objective(sol)8.889281808581273Check the result
The collection stores a reference objective for each problem and modeling, discretized on the default grid of the modeling with the trapezoidal rule: the best objective reached by the configurations of the modeling tested.
reference = OptimalControlProblems.reference_objective(:beam, OptimalControlBackend())8.889281808581256abs(objective(sol) - reference) / abs(reference)The problem page shows the same comparison for every modeling and configuration, and how the JuMP model of the same problem reaches the same objective.
Next
Guided tour: the collection and what each problem comes with.
Solve with OptimalControl: schemes, modelers, solvers, GPU.