## Tuesday, August 06, 2013

### Basketball robot shoots worse than some humans

Today, at the Carnegie Science Museum in Pittsburgh, I watched their resident basketball robot shoot some free throws.  Here's a video of what it looks like.

How accurate do you think the robot is?  Take a guess before you read on.  (Or, just read on -- who am I to give you orders?)

The answer is ... not that accurate.  Well, at least, a lot less accurate than I thought.  When I was there, the robot's FT% was only 83 percent (405 for 488).

I was a bit shocked.  I expected close to perfect.  After all, it's the same throw under exactly the same circumstances, every time.  (Actually, it's two different shots: sometimes the robot throws underhand, and sometimes overhand from behind his back.  But, I witnessed the robot missing shots from both positions.)

It seems wrong, doesn't it, that a human can outperform an expensive robot at a repetitive physical task?  In his career, Rick Barry routinely shot over 90 percent (albeit underhanded).   So, the machine misses almost twice as many shots as Barry.

What's going on?  I don't know.

For what it's worth, here's my theory:

The robot lets the ball roll down the ramp that's his "hand" before actually doing the throw with his "arm".  Maybe the position it reaches varies randomly, based on random differences in friction.  Maybe if a dirty part of the ball contacts a dirty part of the arm, the ball doesn't quite reach the expected point, and the throw misses.

There could be other friction-related issues that cause variation, like, perhaps, the axis of rotation of the ball when it's released.  (The robot hits the backboard every time.)

Any physicists reading who can deliver a more informed hypothesis?

------

If a human can outperform a robot, then it must be that he does *something* better than the machine does.  What?

My guess is: when the human shoots, he can notice if something's a little off, like the ball slips a bit.  In that case, he can adjust his motion on the spot to try to counter that.  The robot, of course, doesn't do that.

That's the only thing I can think of.

-------

I'd love to hear other opinions, because I'm very, very surprised.  I would have bet good money that you could easily make a robot that shoots, say, 98 percent.  Could it really be that tiny differences caused by friction could make such a big difference in outcomes?

Labels: , ,

#### 14 Comments:

At Wednesday, August 07, 2013 10:55:00 AM,  Anonymous said...

Maybe the robot was just "cold" that day? Maybe the robot choked under the pressure when it knew you were watching it? :-) Really, I think you're exactly right with the addition of small variations idea. It could be enough that the position and orientation of the ball in the robot's "hand" prior to shooting is enough variation to cause misses.

At Wednesday, August 07, 2013 12:38:00 PM,  Anonymous said...

Also, assuming the robot is shooting different balls, there could be variations in air pressure/material properties etc that would effect the rebound off the backboard.

At Wednesday, August 07, 2013 1:36:00 PM,  Alex said...

It looks like it lets the ball settle and comes to a complete stop before shooting, so presumably that part is repeatable. The suggestion of different balls seems reasonable if you saw them swap them out, or it could just be use of the ball in general. Was there anything in the museum that could have been causing the backboard to shake, or wind to move the ball?

As a completely different theory, is there a description of how the robot was programmed or built? Maybe there's actually variability built in so that it isn't perfect every time.

At Wednesday, August 07, 2013 10:36:00 PM,  Phil Birnbaum said...

I didn't think that the balls might be different ... that's a possibility! It would still be strange that slight variations in balls (there were three, and they all looked regulation) could cause that big an effect, but it's certainly possible!

I couldn't find anything about how the robot was programmed, but I assume that they tried to make it as accurate as possible. At least that's the implication I got from their signage.

At Thursday, August 08, 2013 12:40:00 AM,  Ben Morris said...

At SSAC there was a booth for a robotics competition in which high schoolers had to design, program and build basketball-playing robots, and those things hardly ever missed.

At Thursday, August 08, 2013 4:27:00 AM,  Anonymous said...

Assuming there is no artificial variability built in. There must be an environmental variability on each trial that causes misses.
If we can exclude changes in wind speed/ direction and outside pressure and further exclude differences in ball pressure, the only thing I can imagine is the shape of the ball.

When the robot throws an underhanded free throw the ball rolls down the rack. Maybe differences in how the ball originally is located on the rack affects the speed of the ball rolling down the rack. That may cause the robot to throw the ball too early or too late.

At Monday, August 12, 2013 10:33:00 AM,  James said...

Maybe the humans who programmed it were bad at their jobs?

At Monday, August 12, 2013 10:04:00 PM,  Phil Birnbaum said...

Good comments, thanks! Maybe Ben and James have it ... the robot wasn't perfected to be 100% ... once they got past 80, maybe they said, "good enough" and quit.

It's still interesting, though ... I would have thought misses were more "macro" in nature, that you have to be off by a lot. It's unlikely that the robot is off by a lot, even if it wasn't designed to NASA standards ...

At Wednesday, August 14, 2013 12:16:00 AM,  Matt O'Neal said...

While the robot's mechanics look simple and very repeatable, it would only take a couple of very slight variations, when run through a multiplicative Monte Carlo simulation to reduce the end results to that low.

Say, the hydraulic pressure in the main piston was 150 PSI for every shot, +/- 1%. That could be reasonable.

Maybe the delta in the three balls air pressure was 3-5 PSI. This difference might not be noticed by us non-basketball players, but might be noticed by the pros, at least subconsciously.

And the orientation of the ball could add another 1% variation. I'm sure all pro ball players have the ball oriented in their hands the same way for every shot- though that may be as much a mental thing as anything else.

All told, variations as seemingly minor as these could compound to result in a final variation of 10% or more.

But then again, while I am a physicist, I am completely guessing here.

At Wednesday, August 14, 2013 2:36:00 AM,  Phil Birnbaum said...

Thanks, Matt. Still seems like a to me lot ... but, that's just my gut.

I wonder if the robot misses with certain of the three balls more than with others? That would help figure it out.

At Thursday, August 15, 2013 8:58:00 PM,  Scott Segrin said...

A basketball is not perfectly spherical. There are the little dimples on it, the seams, the place where they stamp the brand name in the leather, the spot where you pump it with air. Each time you're going to get a slightly different "grip" on the ball. I would assume this to be a big contributor to the variability.

At Tuesday, September 24, 2013 9:46:00 AM,  Anonymous said...

Maybe it takes 10 hours to build a robot that shoots 83%, and it would take 500 hours to make one that gets to 99.9%.

Maybe they stopped when they got it good enough -

At Wednesday, September 25, 2013 7:47:00 PM,  Anonymous said...

Who said robots are perfect? There can be some slop and hysteresis in the mechanics. Just because it's a machine doesn't mean it has no internal variability. There are good robots and mediocre robots.

Ball weights and pressure vary too as does the surface even on one ball, and human can be sensitive to that (a robot could be too if it's fancy enough).

At Friday, October 18, 2013 3:47:00 PM,  Unknown said...

I would love to see those basketball robots play. I've been into basketball sports since I was a kid. I'm into sports betting now and I always win using the experts picks from Dynamitepickspicks.com.