I remember that a few months ago there was a competition in the US where unmaned vehicles (cars and one motorbike as I recall) had to negotiate a 20mile course through the desert. Not one finished the course. Remember, they did not have to contend with Wx, traffic....they operated in 2 dimensional environment. And they did not have to be fail operational.
An event of this sort did occur recently, promoted by DARPA and populated by self-financed competitors from colleges and aspiring small companies - plus a few teams of pub-crawlers. The machines entered were cribbed together on volunteer labor with things one finds 'around the house', so the poor showing of most vehicles was predictable.
Without intending any slight, I think you really sell short the whole concept regieme of autonomous vehicles by relying on such highly visible amateur competitions as proof of current capability.
Please consider the following additional thoughts:
A. Neither military nor industry publish press releases about their "best" stuff unless some great incentive prompts them to do so. Personally I would assume that the state of the art is about 15-20 years ahead of what college engineering students are building as class projects.
B. Nothing about fully implemented autonomous vehicle technology is either simple or cheap. The costs for very limited, but complete & fieldable, systems range from hundreds-of-millions to billions of $$.