rbfx | Game engine with extensive C # support and WYSIWYG editor | Game Engine library

A lot of info from this and you another threads. I will try to handle all as I can.

The problem is not converting pyomo.IndexedVar to float, since when you create the variable in the pyomo model you can do it as a NonNegativeReal. This problem is rather the implicit convertion of not a number (In your code, m.lammda[i,1] is a class, not exactly a number. Even though it has a float value) with the float statement.

As you stated in your comment in another thread, when you try to use the pyomo.environ.IndexedVar into the fitting (interpolation) function Rbf, you raise the TypeError of error trying to convert the an IndexedVar to float (internally in the Rbf function). To evaluate the value you can use model.m.lammda[i,1]() since calling the class returns you the value, but this still will not help, because another error will certainly raise

I recommend you to visit this post to get more information about the potential difficulty of using an external function as a pyomo Constraint or Objective Function

Pyomo got a Piecewise Function Library where you can model a linear (pyo.environ layer also got it in the pyo.environ.Piecewise) or multilinear interpolation, but you will need to work with the kernel layer, it use scipy to generate a Dlunay for the interpolation, you can check using help(pyomo.kernel.piecewise_nd). This surely may work to you, but it requires a little be of work.