To actually deliver a physical, one-of-a-kind, working vehicle of that scale & novel function, for low-digit millions, seems pretty impressive to me - especially if the effort documented enough of their choices/discoveries/limitations to help inform later efforts.
DARPA R&D money always feels like money better spent than a lot of other military spending. We get future dividends, often civilian.
He argues the same point about the states roll in RnD (also explicitly calls out DARPA) and also their role in booms and busts.
https://en.wikipedia.org/wiki/Defense_Production_Act_of_1950...
[1] as far as I've seen, soviet vending machines used a shared glass. They must've been very confident both (a) that the glasses wouldn't grow legs, and (b) in their public health initiatives.
Whether the US Army needs 9000 tanks is a separate discussion though.
If you're looking at needing to fight China and Russia, at the same time, over an extended period of time (2+ years), having 9,000 tanks in your back pocket seems like a good idea.
Also, it's a jobs/pork barrel program. Congress keeps ordering tanks that the Army puts into storage because the M1 is a beast of a machine that keeps a lot of people employed. The US manufacturing base is pretty depleted, and ordering tanks you don't necessarily need is a good way to keep the skills and tool chains you could need in the future around.
Sure, that's nothing new to anyone who's familiar with Cold War economics. The Soviet Union was able to keep its economy crawling along without sanctioning free enterprise by making enormous investments into collectively-owned (i.e. state-owned) products like nuclear submarines and combat aircraft. Turns out you can achieve much the same goal but also benefit the people if you put that effort into state-funded infrastructure. Whether the US body politic will support reducing M1 Abrams production in favour of fixing the USA's crumbling infrastructure is again a separate discussion.
More precisely: for the 2020 fiscal year, the US Army did not procure any new M1 Abrams tank, and is not budgeting any for the 2021 year either. Instead it is modernizing existing tanks at a cost of $1.7 BN for 165 units in '20, and $1 BN for 89 units in '21. This comes at about $10.6 MM per tank.
[1] https://www.asafm.army.mil/Portals/72/Documents/BudgetMateri...
[2] https://www.asafm.army.mil/Portals/72/Documents/BudgetMateri...
[1] https://archive.macleans.ca/article/1985/11/4/the-pentagon-f...
[2] https://www.inflationtool.com/us-dollar/1981-to-present-valu...
A two ton diesel powered hexapod.
I was curious; the Manhattan Project cost $2 billion (in 1939) and employed 130,000 people (https://en.wikipedia.org/wiki/Manhattan_Project, first paragraph).
Planning foot placement is a bit harder but I would say the compute heavy part today would be perception.
But figuring out where to place the feet, which I consider a perception problem, is very hard. But I think that’s what you’re talking about and I just use different terminology. That’s why I was confused. Sorry if I came off wrong.
Given the lack of research, but perhaps the high experimentation, millions make sense.
Building the prototype at full scale, not so much. If they'd gone through a few iterations starting with a dog sized version, it might have worked out better. But there was a political problem. DARPA funded a pony-sized legged walker in the late 1960s[2], and it was publicly shamed as a waste of taxpayer money by Sen. Proxmire. So DARPA went for one big enough to carry a real load.
Early thinking about legged locomotion was all about gait and foot placement. It turns out that balance dominates foot placement, and slip control dominates balance once you get off the flat.
Today, most of the problems have been solved, and there's still no market. Which is disappointing.
[1] https://youtu.be/coNO9FpDb6E
[2] http://cyberneticzoo.com/walking-machines/1968-phony-pony-fr...