While I've thought about this before, and in each year our team increased the amount of autonomy involved, the problem is in the limitations posed by FIRST.
Too little money. Too little time. Too few motors. Too little battery. Too small a processor.
If all that were not an issue, then I have no doubts that a fully-autonomous robot would work well.
The problem is that FIRST allows you four of a really powerful computer when you run your robot, a computer than can react to all sorts of stimuli - vision, sound, the score of the match. It has built in feedback controls and is powered by coffee and donuts.
Four living, breathing, thinking people are tough to beat.
