Re: Welcome to DARPA-Land where budgets never get cut…
Factual corrections time:
DARPA cuts projects midstream when they don't meet the metrics. Ex: the Airlander. (You didn't know that the Airlander was a failed DARPA project?)
DARPA terminates all projects after several years. If they were good then someone else will continue them.
Carnegie Mellon (and other) scientists did the initial research before the Army; see my previous point.
The first grand challenge was a 200 miles course in the desert; the best team failed about 7 miles in. The 2nd running was 130 miles; 5 teams finished.
DARPA did not pay for the research, but awarded a $1M prize after the fact to the winner and nothing to the others. This seems an optimal use of taxpayer money.