As we have no definition of the word "moral" that withstands any scrutiny, and as we appear to want our machines to be perfect while accepting the fact we're totally useless at difficult tasks like driving, perhaps the easiest resolution to the so-called "ethical problem" of autonomous vehicles is simply to program them to do what humans do in equivalent circumstances: freeze and plough straight ahead into whatever it was the vehicle was last looking at when the situation occurred. And then report something like "the tree jumped right into the middle of the road!"