First consider a simple polymer, like polyacetylene:
Its torsional barrier can be determined by DFT calculations by considering "enough" units and performing a simple dihedral scan. "Enough" appears to be 5-7 units (J. Chem. Phys., 2014, 140, 054310)
The same method does not work, because the structure is too flexible. Different values are obtained after multiple turns around the central unit. Is it possible to use molecular dynamics to find the effective/average rotation barrier?
My guess would be that we could simulate this polymer over a certain period of time, then make an histogram of the central dihedral angle. We could then calculate the torsional barrier by looking at the relative frequencies of the angles (e.g. if we get a flat distribution, the barrier is 0, if some angle is very disfavored, the barrier should be high. I'm not sure how to quantitatively calculate the barrier though)