To be precise, I'd have to have the data. And no, I won't come up with something perfect, just something a lot better than what we have now.
I don't think it's a conceptually hard problem. Predict performance in college based on socio economic factors. Colleges should have plenty of relevant data. Adjust admissions policy to select those with highest predicted performance. That's if we're trying to take the people with most potential. Probably a combination of IQ, grades, and parental economic status, along with population statistics for your high school cohort.
If we just want to help the disadvantaged, come up with measures of disadvantage, and apply. Parents wealth/income is probably the place to start, and would avoid the absurdity of having the President's daughters be classified as disadvantaged. One shouldn't want to help the most disadvantaged, because they're unlikely to benefit. Likely the same factors as before, though changing the goal to helping the disadvantaged will involve a trade off between likely achievement and disadvantage.
I don't think it's a conceptually hard problem.
As I said in my response to Viliam_Bur, it's easy to say "find out what the relevant factors are, address those," much harder to detail exactly how one works out what the significant factors are, let alone how to best address them.
Keep in mind that sorting by IQ is still a strong political taboo, and more politically viable proxies, such as SAT scores, are liable to be corrupted by factors such as SAT prep. Plus, target minorities are subject to an IQ gap which may be due to some of the cultural ...
Related: Heuristics for Evaluating the Soundness of the Academic Mainstream, Admitting to Bias, The Ideological Turing Test