There is this question here: Recommended way to deal with period Minuit parameter but no answer either.
Maybe this trick works:
- If you know that the result of your parameter is likely to be in the area around 0 radians (2pi), then replace your parameter z in radians with z = arcsin(x), and fix the limits of x accordingly. This way the boundary will be somewhere else
- Likewise if you know that the likely result will be at pi/2, you can change to arccos(x)