Having a robot hit a baseball is trivially easy given current technology. You just make a robot that swings the bat incredibly fast. If the bat speed is several times that of a human’s, it won’t matter much what the ball’s break is.
I am astonished anyone would think this difficult.
The reason a human has trouble hitting a curveball is because there’s a time difference between the decision to swing at a predicted point of contact, during which the ball has time to move from the assumed point of contact to a different one. Since you can only swing the bat so fast, you must make that decision at a certain point in the ball’s flight and commit yourself to aiming the bat at a particular point in space. If you incorrectly guess the ball’s path, the bat will not arrive in the correct space at the correct time, and beyond a certain point it is too late to do anything about it.
To defeat breaking balls, construct a robot that makes the decision to swing ten times faster than a human, and actually swings the bat five times faster than a human. Problem solved. The decision to swing can now be made much later in the ball’s flight, and the predicted point of contact is little different from the actual point of contact, even if the robot can be “fooled.” It’ll be fooled by a millimeter.
Want a baserunning robot? No problem. Just build a robot that goes 75 miles an hour. It’ll make Cool Papa Bell look like Cecil Fielder.
Fielding robot? Again, no problem; just make it ultra fast, and it will not have to have the reaction times that make excellent outfielders what they are. An outfielder needs fast reaction times because her/his speed is limited - a very, very fast runner can only do 25 MPH. A robot could do 80. So it can wait until it’s sure where the ball will be.
This is all quite doable. You just don’t have the robots look or act like humans. I don’t know what the point would be, but it’s easy enough to engineer, assuming you have a lot of money.
Thing is, there does appear to be people who are working on this, but I haven’t seen any examples yet of anything better than this robot hitting balls pitched to it from another robot at 25 mph. Now, that article is from 2009, but I have not seen any updates yet. I suppose it could be for lack of interest in funding a project like this, but there does appear to be a few people working on this, and nothing has surfaced quite yet. While it seems like a fairly easy problem to me, or rather, not as monumentally difficult as other posters here are suggesting, I wonder if the devil’s in the details and it’s a bit computationally more difficult than we are giving credit for it being.
I’m a little astonished that anybody would think it’s not… it’s not just a question of swinging the bat fast; the robot needs to judge from viewing the pitcher where in the strike zone the ball will be, (if it’ll be in the strike zone at all.)
It may be a solved problem given current technology–but that doesn’t mean it’s easy, just that the hard work’s been done.
Well, I was assuming the fact we’re standing on the shoulders of giants and the funding is in place. Lots of things are not major engineering challenges, but they still cost a lot of dough.
I think making a human-shaped robot that can hit is probably not that hard. Making a robot that can crush a baseball, sure. That is a very specific type of motion that is well understood and you could make a human-shaped robot to do that.
But to make the same machine now be able to run the bases… wow, that’s way hard. As has been pointed out, just getting robots to walk and run is a recent thing. To have a machine that can do both is probably not something that exists now.
To now teach it to play second base… well, now we’re well into science fiction territory.
Farnsworth: He’s good all right. But he’s no Clem Johnson. And Johnson played back in the days when steroid injections were mandatory.
Bender: Clem Johnson? That skin bag wouldn’t have lasted one pitch in the old Robot Leagues! Now Wireless Joe Jackson, there was a blern hitting machine!
Leela: Exactly! He was a machine designed to hit blerns! I mean come on! Wireless Joe was nothing but a programmable bat on wheels.
Bender: Oh and I suppose the Pitchamatic 5000 was just a modified howitzer!
Leela: Yep.
If you’re going to give the batting robot the edge because it has more time to make its decision, you have to take into account the pitching robot whose fastball might be ~250 mph.
Can you point to the specific existing technology that you believe shows it’s trivially easy to trigger a bat swing dramatically faster than a human’s on a narrow enough adjusted swing path?
Also curious as to how closely you follow the engineering and robotics literature.
If I provided an example of a machine that was actually MORE impressive than a baseball-hitting robot - let’s say a robot that could identify an incoming projectile from thousands of meters away, determine its speed, trajectory and other characteristics of flight, and then calculate the correct timing to not only strike it with another object, *but to do so with another flying object *- as if throwing a baseball at another baseball and hitting it in mid-flight - would that suffice to demonstrate the technology exists? What if I said they’d had stuff like this for decades?
I dunno, maybe it’s just me but I’d have a hard time getting worked up or excited about rooting for the robots from Detroit playing the automatons from Boston
The most amazing thing about that clip was how easily the robots stood back up after they fell, and how natural it looked. Much more than their walking and kicking the ball. Nice to see no diving or cars as well (though I didn’t watch it all).
Wouldn’t this be a fairly trivial application of PitchFX? Given the velocity, spin rate and angle, coming up with a projected trajectory shouldn’t be too hard. You might have to adjust for local atmospheric conditions, but once the robot has seen a couple of pitches, they can probably approximate that too.
Of course, this all goes out the window if the pitcher is throwing a knuckleball.
The speeds and reaction times are comparable, though. You can’t see a baseball coming from a mile away, but it’s also not capable of supersonic speeds and taking multiple evasive action maneuvers.
The reaction time needed to hit a baseball is super fast to a human, but not fast to a machine. Machines are good at making simple decisions very, very quickly.
You need to watch the next video (at least for me it came up next) of the best dives and falls of the 2016 robot soccer. I was literally in tears watching it :D:D
With the dangers of American football becoming more apparent, maybe I should have asked about football playing robots for the NFL instead.
RickJay: the Phalanx is pretty impressive, so is Iron Dome, but while hitting a baseball requires similar technology, can the systems process the information fast enough to calculate a solution and make contact, not with a stream of bullets, but with a single wood stick that it only gets one chance with? To see where I am going, let’s say the Phalanx is going to engage an Exocet missile. If Wikipedia is accurate, the Exocet has a minimum range of 43 miles and a top speed of 705 mph. Phalanx has an effective range of about 2 miles. If doing the math correctly, it has a bit over 3 minutes to identify the target, project an intercept, and engage. Can we get that time down to less than a second in the batter’s box?
Rysto: Is PitchFX accurate enough? Being off by a little on location or speed could throw off the swing and be the difference between a hit or a strike.
