How far are we from baseball playing robots?

Great Debates
1

“This is a simple game. You throw the ball. You hit the ball. You catch the ball.” Skip - Bull Durham

I was watching my Astros play today and got to thinking about how, at least in the 80’s, folks thought robots would be pretty limited in their use because they would only be built to do very specific repetitive tasks (think the robots building cars). Since that time we’ve made some great strides in sensors and software so I got to thinking, how soon could I field a robot baseball team?

As humans, we are pre-wired for throwing, hitting and catching, in fact much of the thought process to accomplish these actions seems to take place instinctively. But teaching robots to play these games requires an integration of sensors and software that we are only just starting to develop. Lets take a look at what are robots would need to be able to do:

Throwing the ball seems pretty easy. It’s simple ballistics, but with variables that software can already compensate for (I think). Deciding where to throw, that’s a whole other issue.

Hitting a baseball is considered one of the hardest things to do in sports. In the major leagues, a batter has something like 225 milliseconds to decide and then begin to swing at a 90mph fast ball. Then he has to be accurate to something like 7 milliseconds for his swing to put the ball in play, not to mention managing the vertical and horizontal axis of the swing to make contact with a 3" diameter ball using a 2 1/2" diameter bat. Millimeters can make the difference between hit or a pop-out.

In the field, the fielder has to locate the ball, predict a path, plot an intercept point, and make whatever adjustments are necessary during the flight time, and, finally, make the catch.

The question for discussion, do we have the tech to build self-contained robots that could autonomously accomplish these actions? What challenges do we still face?

(I left out base-running, as I think we already have the technology to get robots to run the bases. We have self driving cars, for crying out loud! Then realized there are still a ton of decision that go into base-running, too. However, mentioning it above would have messed with symmetry of the nice set-up from the Skip. :slight_smile: )

2

We’re a long ways away, even for just running. Getting a 2-legged robot to walk without falling over is actually rather hard, running would be harder still.

Since this is GD and not GQ, I’m going to ask: Why would we want baseball playing robots? It seems to me that sports is one type of activity that we don’t want robots taking over.

ETA: Despite the quote from Bull Durham, baseball is not a simple game.

3

I’m no expert, but judging from this footagefrom the 2016 Robot Football world cup, we’re decades away from anything that could play baseball with standard rules and equipment.

4

I wasn’t even thinking of trying to make the robots human shaped, although they should be roughly human-sized. Two-legged locomotion is a pretty challenging of itself, but using wheels are just fine. You can even change the robot’s equipment out (human players do the same) so you could have a fielding package and a hitting package to mount on the movement base.

The RoboCup footage was interesting. I’m actually pretty impressed that the robots could move without constantly falling over and even get back up unassisted.

5

There have been a couple of suggestions that this would do better in the Game Room. However, at the moment it still looks more like an engineering discussion than a sports discussion, so, for now, I
am leaving it here.

6

Making the required decisions is not hard at all, and I have a hard time imagining that even an amateurly-programmed robot would do worse than a human. When you’re making decisions for others to react to (like what kind of pitch to throw), your goal is unpredictability, for which you can’t do any better than randomness, at which humans suck. And when you’re deciding how to react to what others have done, it’s mostly just straightforward distance-velocity calculations, like calculating whether you can get to base before the thrown ball does, and those are trivially easy for a computer.

7

Assuming it could someday be possible, it would be like a self-sustaining video game with no human input. Would they actually need physical balls or bats? Or an actual field? And who would watch it, other robots?

8

I imagine it would be a useful tool in the development of human baseball skills.

9

Certainly much more interested in the engineering challenges we still need to overcome, not so much the game side of things.

I am not so worried about the robot looking like a human out there on the field. The outfielders could be wheeled metal cans with ball return launchers of some kind. It’s the finding the ball as it comes off the bat, projecting its course, and then moving to catch that I’d like to know where things stand.

Same with hitting. The robot can be a metal pole with a bat and a radar station on top, what challenges are still out there before it can be a .200 hitter against a human pitcher? We can track small warheads in flight, so I wouldn’t think tracking a baseball is that difficult, we can probably predict where the ball will go. Can we do that in the 225 milliseconds we need to decide whether to swing? And could we even track the bat and ball well enough to connect?

10

Sports is a form of entertainment and I don’t see us anytime soon finding baseball playing robots more entertaining than baseball playing humans.

Battlebots haven’t done away with Boxing or MMA.

11

It might be a cute novelty, but only science nerds would become hardcore fans. Baseball love is all about emotional nostalgia and regional loyalties.

12

Who knows? With time and no distractions, somebody may one day manufacture a team of automated baseball players. But could he make them have fun while they throw, hit or catch? I doubt it.

13

A lot of baseball is a psychological duel between the pitcher and the batter. How would robo ball replicate that? How would it replicate these? Roboball is an interesting engineering concept, but it would be a dull as crap game.

14

Back in the 80’s (that’s 1980s, NOT 1880’s - I’m not THAT old) some folks in Japan had too much time with a computer.
They calculated Pi (the 3.14159 thing) to some ridiculous number of decimal positions.
Then the program ended. It never recorded - print or magnetic the result, It didn’t even display it.

We can write programs to “play” baseball or football, or bowling, or making cherry pies.
A printer typing out (keep an impact printer, just for the hell of it) “Yankees -5 Dodgers -3” would be exactly as interesting as watching “robots” launch balls, swing a stick at the ball, etc.

“Kinda neat - glad I saw it ONCE” would be my response to “Robotic Baseball Teams”.
There will be somebody, somewhere to do this, just as somebody calculated Pi to several hundred positions.
But Why?

15

Rod Serling explained playing robots vs. humans,

16

Pitching? Sure. Hearing the crack and knowing from that where the ball is going? Difficult. Hitting a curveball? No way in my lifetime, nor my next.

17

Apparently it’s already happened.

18

I find this relevant. Robot Apocalypse

19

See also John Bellairs’ book The Eyes of the Killer Robot.

20

I would think the curveball problem would be a much easier than fielding fly balls (if that’s what you mean by “knowing where the ball is going” – that is, not simply predicting the path of the ball but actually fielding it.) I would certainly expect a robot to be capable of that (hitting all types of pitches including breaking balls) within my lifetime.