If you use Coulomb friction model you can set dependency of friction coefficient not only on the default variables (slip rate, pressure and temperature) but also some field variables:
*FRICTION, DEPENDENCIES=n
u, y, p, T, f
where: u - friction coefficient, y - slip rate, p - pressure, T - temperature, f - field variable.
Subroutines FRIC_COEF/VFRIC_COEF correspond to this model.
I think that you could use USDFLD (or VUSDFLD in Explicit) to relate this field variable to strain. However FRIC/VFRIC subroutine used when Coulomb model is not enough should also support the use of solution-dependent state variables.