Pyomo，TypeError：'numpy.ndarray' 对象不可调用

Question

I am trying to use scipy commands in Pyomo , an optimizer based in Python.

My goal is to develop an optimal trajectory for aircraft in the presence of a wind field. I have data measurements at grid points. I have used scipy to generate the interpolation as

xi, yi = np.linspace(X2.astype('float').min(), X2.astype('float').max(), 100), np.linspace(Y2.astype('float').min(), Y2.astype('float').max(), 100)
xi, yi = np.meshgrid(xi, yi)

# Interpolate
rbfX = sp.interpolate.Rbf(X2, Y2, Wx2, function='multiquadric')
zXi = rbfX(xi, yi)

being sp how I call scipy. This is the way I have to calculate the speed of wind at any point. I am interpolating the windspeed.

Then, inside of the pyomo part, I write

def Wind_lammda_definition1(model, i):
    return m.Wind_lammda[i,1] ==zXi(m.lammda[i,1], m.phi[i,1])
m.Wind_lammda_const1 = Constraint(m.N, rule = Wind_lammda_definition1)

being m.lammda and m.phi the position of the airplane.

Unfortunately, once I run the code, I get rhe following error:

    return m.Wind_lammda[i,1] ==zXi(m.lammda[i,1], m.phi[i,1])

TypeError: 'numpy.ndarray' object is not callable

I have checked this and this threads and as far as I see, this is an error message which appears if there are syntax errors. However, I haven't found any and I think that this happens becasue I cannot port scipy into Pyomo. Is this true or can it be fixed?

Answer 1

Let's setup a Rbf object:

In [177]: X,Y = np.meshgrid(np.arange(4),np.arange(4))
In [178]: Z = np.sin(X-Y)
In [179]: fn = interpolate.Rbf(X,Y,Z)
In [180]: type(fn)
Out[180]: scipy.interpolate.rbf.Rbf

That's like your rbfX . According to the docs it is callable , that is it has a __call__ method.

If we call it with 2 scalars, we get a single element array:

In [181]: fn(1.5, 2.3)
Out[181]: array(-0.73094599)

Call it with a pair of arrays, we get an array with matching size:

In [184]: fn(np.arange(1,3,.5), np.arange(0,2,.5))
Out[184]: array([0.84147098, 0.85639014, 0.84147098, 0.85639014])

This is what your zXi . Look at its type , shape , and dtype . It's an array. It cannot be "called", and there are specific rules as to how it can be indexed.

zXi(m.lammda[i,1], m.phi[i,1]) is wrong because it tries to "call" a numpy array. But zXi[m.lammda[i,1], m.phi[i,1]] apparently has invalid indices, quite possibly floats. I haven't examined what they are.

rbfX(m.lammda[i,1], m.phi[i,1]) might work if the 2 arguments are meant to be interpolation points.

zXi = rbfX(xi, yi) creates an array (probably 2d (100,100)) that contains values interpolated at the xi,yi mesh. You can pick an element, eg zXi[50,50] or array of values. But you can't treat it like the interpolation function, giving it float values.