What would be some good rules for a robot golf tournament? And how many years before a golf-playing robot could beat a pro?

August 14, 2012 6:24 AM

What would be some good rules for a Robot Golf tournament? And how many years can we expect to wait before a golf-playing robot could beat a pro?

The more I think about it, the more the world could use a robot golf competition.
So, given the state of current technology, how well could a robot play golf?

1) From what I've seen online, there are a few bots designed to drive and putt golf balls... but has anyone ever attempted to have a bot play a full hole?

2) If a golf tournament were to be held for robots, what sort of rules would need to be established? What level of autonomy could be expected of them in the first competition?

3) The DARPA Grand Challenge was awesome, and there we saw bots go from failure to mastery in just a few years- could we expect the same thing for golf-playing robots?

4) Finally, how many years (decades?) do you think it will be before a robot could beat a pro on a full course?

Thanks for indulging me on this question... because seriously... golf playing robots would be totally awesome.
To me all of this doesn't seem to be a question of if it will happen (because certainly it will, how could it not?), but rather when.

posted by mnop to technology (11 answers total)

I think you have to call the league the RPGA.

Since the robots are playing golf, not merely getting balls into holes, the robots must use standard golf clubs. Clubs may not be permanently attached to the robot. They are allowed to self-caddy however.
posted by shothotbot at 6:32 AM on August 14, 2012 [1 favorite]

It would seem appropriate to take the long view and start with rules that enforce limited autonomy, and then plan for the rules to tighten over time.

For example, the earliest tournaments might allow for the bots to be positioned by humans and only be responsible for their swings. Later tourneys would require the bot to approach independently.

One might also want to consider rules that limit the sort of meteorological devices that such a bot can use. A bot that can precisely measure heat, humidity, barometric pressure and wind speed would definitely have advantages over human players.

Also, it seems appropriate to mention this Penny Arcade strip
posted by DWRoelands at 6:32 AM on August 14, 2012 [1 favorite]

The Rules of Golf disallow the use of any sort of rangefinding/distance-measuring equipment (Rule 14-3) So, to be legal, a robot couldn't have any sort of distance-measurement capability that exceeded the accuracy and ability of human eyesight. You'd actually have to engineer-down the "vision" of the robot.
posted by Thorzdad at 6:39 AM on August 14, 2012

What level of autonomy could be expected of them in the first competition?

To start with, go for miniature golf -- a nice predictable surface with interesting obstacles -- and use a normal rulebook. Require that a robot do everything just like a human competitor -- require that the robot find the hole, line up the shot, hit the ball, track the ball, recover from misses, attack and destroy other robots, etc. -- and award points for each such requirement it meets per hole. If you have help it do some task (find the ball, line up a shot, etc.), the robot doesn't get points for that task. The only thing you might help it with as a matter of course is moving from hole to hole, assuming there is no rollable route between holes.

The Rules of Golf disallow the use of any sort of rangefinding/distance-measuring equipment (Rule 14-3)

The rules of golf allow you to use your eyes and your brain to estimate ranges. For a robot, that's whatever you give it. It should even be allowed to construct a course map during previous games, because everyone else gets to remember courses they've played on.
posted by pracowity at 6:49 AM on August 14, 2012 [1 favorite]

I would assume this would include any sort of GPS tracking. Or, at least, the robot could not use GPS to do anything other than identify its current position on the course, but not correlate that position to any other measurement, thus, in effect, measuring exact yardage.
posted by Thorzdad at 6:50 AM on August 14, 2012

The rules of golf allow you to use your eyes and your brain to estimate ranges. For a robot, that's whatever you give it. It should even be allowed to construct a course map during previous games, because everyone else gets to remember courses they've played on.

I suppose the devil is in the details. At what point does visual estimation become exact measurement? As I said, in order to conform to the Rules, you would probably have to dumb-down the robot's abilities.

This is, of course, assuming the intent is to have to robots adhere to the Rules of Golf, in order to achieve as close to an apples=apples comparison. If the point is to simply get robots playing "golf", then I guess the Rules are a moot point..
posted by Thorzdad at 6:54 AM on August 14, 2012 [1 favorite]

@Thorzdad I think the intention at the start is to just get them playing, then as time goes on perhaps have them ahdere to more rules as their skills improve. Still though, robots using their super robot eyes is part of the appeal of having robots playing golf.

@shothotbot RPGA sounds good, and using real golf clubs seems like the way to go. They can self-caddy, but I assume they can have a real caddy as well? BTW, Hotshot Bot would be an awesome name for a golf-playing robot.

@DWRoelands That's what I think would have to happen as well... more human-assistance at first (like people conveying their bots roughly to the ball's position each time), but then as their improve require more autonomy. And that's a great Penny Arcade strip, how'd I miss that?

@pracowity minigolf might be the right way to start all of this. I've seen videos of bots playing minigolf-like games in the lab, but has it ever been tried on a real course?
posted by mnop at 7:10 AM on August 14, 2012

i'd want the robot to use as much rangefinding equipment as possible. it would be very satisfying to see a robot hit holes in one for an entire 18 hole course
posted by cubby at 7:20 AM on August 14, 2012

I think the point is for the bot to eventually be better than a human. That's going to be more difficult than it sounds, I think, even with meteorological sensors, range finding, etc. An experienced pro's "gut feel" is not just random guesswork.

I'm thinking the bot should have to be (eventually) self-autonomous from loading the set of clubs. Rather than start on a mini-golf course, I'd probably start at the driving range, just making the robot capable of addressing the ball, selecting a club, and making a shot.

But thinking of all the things the bot would have to do: wow. Even allowing GPS and a map, it would have to track the ball, transit to it over uneven terrain, evaluate the lie...

That would be super-difficult, recognizing and dealing with a poor lie autonomously. Tree in the club swing path? Obstacle in the intended ball flight path? Ball in the gravel or sand?

Plus, you would have to get all that gear on a vehicle that wouldn't leave track marks all over the green.
posted by ctmf at 7:38 AM on August 14, 2012

I think there is enough randomness involved in the game of golf where a pro could sometimes still beat a robot that played technically perfectly.
posted by ctmf at 7:40 AM on August 14, 2012

In robotics, from what I understand, a large factor is weight. The heavier you can build your robot, the more stuff you can put on it, and, in general, the 'better' it is. A lot of the cutting edge stuff is in miniaturization, better programming and optimizing, and AI decision trees.

So, do the robots need to make a human style swing motion? Or can they just aircannon the ball?

Must the robot be bipedal, or can it have treads or wheels?

Must the robot 'see' the ball, or can it home in on radio/etc?

Must the robot 'see' the hole, or can it home in?

With the right limitations set (IE, probably few and a lot that benefits the robot) it would probably be possible very soon- 5 years or less. Arobot playing like a human has to would probably be much harder.

Sadly, DARPA is well funded by military types and has some ulterior motives... Robogolf, while cool, doesn't, which lengthens the design phases a lot.
posted by Jacen at 10:39 AM on August 14, 2012