"Great! Now what about amounts? Does 2 adults trump a child?"
Not sure why you bring up things like the Queen, given that's already covered by the "not discriminating against anyone" part - if it's a choice between hitting the Queen or hitting a single other human, the car isn't allowed to choose based on the fact that one of them is the Queen. But as you note, there are an awful lot of situations that aren't a simple 1 vs. 1 choice, and automated cars will inevitably be required to discriminate between who they kill based on some criteria. Blindly legislating that they're not allowed to make a choice over who to save can't prevent situations arising that will require exactly that.
Of course, the bigger problem is that this isn't a question people agree on in the first place. It's all just variations of the Trolley Problem - a runaway train is heading towards 5 people on a track, if you pull a switch you can divert it to a different track with only 1 person standing on it; do you pull the switch or not? In this bare problem, most people will say you should pull the switch, but not all even agree on that. Throw some complications into the mix and things get, well, complicated. Basic complications such as changing the number of people or saying the lone person is your partner, the 5 are clones of Hitler, and so on, can be relatively easy to deal with if you agree on the base problem, and the latter two are the sorts of things that would be covered by the recommendations given here. But things like removing the lone person from the track and instead saying you have to personally murder them in order to reach the switch can make huge differences in how people view the problem, and that's exactly the sort of scenario that's most analogous to the issues autonomous cars will face.
Long story short, we're trying to decide how autonomous cars should behave without actually agreeing on how humans would or should behave in an identical scenario, and in the absence of a universally agreed objective morality, it's not going to be an easy issue to solve.