Seemingly Unrelated Regression vs Maximum Likelihood ~ best for constraints?

I am currently doing research and I am utilizing an almost ideal demand system. It is essentially a regression equation that seeks to model consumer demand. It has prices and expenditure as independent variables and budget share (how much of a good that is consumed, ie, 30% of your budget goes to food) as the dependent variable. I have seen papers utilize maximum likelihood and iterated seemingly unrelated regressions because the model has coefficient constraints. Which model would be best for working with constraints? And also how exactly does maximum likelihood get applied to constraints? My understanding is that ML seeks to estimate parameters based on a small sample, but never heard anything in my years of schooling about it applied to constraints. Any help would be much appreciated!