MTK example no longer work #116
Comments
|
@baggepinnen I just tested and the example work well with MTK v9.41. It stopped working with v9.42. Do you see anything v9.42 release that could have break this code ? Should I open an issue on MKT github ? For now I'll just restrict compat to 9.41 in the documentation
|
|
@AayushSabharwal did anything change in parameter handling or otherwise for observed functions recently? |
|
Yes, but it shouldn't cause that error. If you give me something I can run, I'll pinpoint what broke |
|
This is the example |
|
Here's a copy-paste of the code in the manual: using ModelPredictiveControl, ModelingToolkit
using ModelingToolkit: D_nounits as D, t_nounits as t, varmap_to_vars
@mtkmodel Pendulum begin
@parameters begin
g = 9.8
L = 0.4
K = 1.2
m = 0.3
end
@variables begin
θ(t) # state
ω(t) # state
τ(t) # input
y(t) # output
end
@equations begin
D(θ) ~ ω
D(ω) ~ -g/L*sin(θ) - K/m*ω + τ/m/L^2
y ~ θ * 180 / π
end
end
@named mtk_model = Pendulum()
mtk_model = complete(mtk_model)
function generate_f_h(model, inputs, outputs)
(_, f_ip), dvs, psym, io_sys = ModelingToolkit.generate_control_function(
model, inputs, split=false; outputs
)
if any(ModelingToolkit.is_alg_equation, equations(io_sys))
error("Systems with algebraic equations are not supported")
end
nu, nx, ny = length(inputs), length(dvs), length(outputs)
vx = string.(dvs)
p = varmap_to_vars(defaults(io_sys), psym)
function f!(ẋ, x, u, _ , p)
try
f_ip(ẋ, x, u, p, nothing)
catch err
if err isa MethodError
error("NonLinModel does not support a time argument t in the f function, "*
"see the constructor docstring for a workaround.")
else
rethrow()
end
end
return nothing
end
h_ = ModelingToolkit.build_explicit_observed_function(io_sys, outputs; inputs)
u_nothing = fill(nothing, nu)
function h!(y, x, _ , p)
y .= try
# MTK.jl supports a `u` argument in `h_` function but not this package. We set
# `u` as a vector of nothing and `h_` function will presumably throw an
# MethodError it this argument is used inside the function
h_(x, u_nothing, p, nothing)
catch err
if err isa MethodError
error("NonLinModel only support strictly proper systems (no manipulated "*
"input argument u in the output function h)")
else
rethrow()
end
end
return nothing
end
return f!, h!, p, nu, nx, ny, vx
end
inputs, outputs = [mtk_model.τ], [mtk_model.y]
f!, h!, p, nu, nx, ny, vx = generate_f_h(mtk_model, inputs, outputs)
Ts = 0.1
vu, vy = ["\$τ\$ (Nm)"], ["\$θ\$ (°)"]
model = setname!(NonLinModel(f!, h!, Ts, nu, nx, ny; p); u=vu, x=vx, y=vy)
mtk_model.K = defaults(mtk_model)[mtk_model.K] * 1.25
f_plant, h_plant, p = generate_f_h(mtk_model, inputs, outputs)
plant = setname!(NonLinModel(f_plant, h_plant, Ts, nu, nx, ny; p); u=vu, x=vx, y=vy)
α=0.01; σQ=[0.1, 1.0]; σR=[5.0]; nint_u=[1]; σQint_u=[0.1]
estim = UnscentedKalmanFilter(model; α, σQ, σR, nint_u, σQint_u)
Hp, Hc, Mwt, Nwt = 20, 2, [0.5], [2.5]
nmpc = NonLinMPC(estim; Hp, Hc, Mwt, Nwt, Cwt=Inf)
umin, umax = [-1.5], [+1.5]
nmpc = setconstraint!(nmpc; umin, umax)
using Plots
N = 35
res_ry = sim!(nmpc, N, [180.0], plant=plant, x_0=[0, 0], x̂_0=[0, 0, 0])
plot(res_ry) |
|
The observed function expects an |
|
That looks like a bug in the observed function building then, the system |
|
Right, I didn't notice the |
|
This is because prior to calling mtk_model = complete(mtk_model)Which doesn't have |
|
I can make sys1 = ODESystem(...)
sys2 = deepcopy(sys1)
sys1 = complete(sys1; split = false)
sys2 = complete(complete(sys2); split = false)This is because |
👍
Is that likely to be a concern? |
|
Well a common workflow is |
|
we take care in the example above to use the parameter order defined and returned by |
|
Yeah that should be fine |
|
@baggepinnen does that means the the problem should be solved on the latest stable MTK release ? (V9.49) |
|
I just tested on MTK master and it worked. I think MTK might need a new release before it works, @AayushSabharwal? |
|
I'll tag MTK shortly, just waiting for CI to pass on master. It's currently queued |
@baggepinnen The MTK example in the manual stop working and I'm not able to pinpoint why. It was working well when I released v1.0.1. It now crashes inside the
h_function (the function returned byModelingToolkit.build_explicit_observed_function(io_sys, outputs; inputs)with the stacktrace:Do you have any idea why ?
edit: Also, what's your trick to debug these kind of error? I looked inside the lines at
[1],[2], and so on in the stacktrace but the code is "hidden" because of the macros and stuff.The text was updated successfully, but these errors were encountered: