PDA

View Full Version : Canadian pool-playing robot



dr_dave
12-05-2005, 08:03 AM
A guy in Canada (Dr. Michael Greenspan, at Queen's Universtiy, Ontario) is developing a pool-playing robot, complete with automated computer vision systems and software for aiming and strategy. They have a long way to go before Efren will feel nervous, but the preliminary results are pretty impressive (for a machine anyway). A good overview video (http://www.exn.ca/news/video/exn2004/10/20/exn20041020-poolbot.asx) (created by the Discovery Channel) can be viewed online. There are some additional clips on the creator's website (http://www.ece.queensu.ca/hpages/faculty/greenspan/index.html), under the photo.

Check it out. It's quite interesting.

Regards,
Dave

dr_dave
12-05-2005, 02:37 PM
How long do you guys think it will take for somebody to design a pool-playing robot that can beat the best human pro?

If you watched the videos, one thing that is obvious is that the Canadian robot "breaks like a girl." This is a major liability, but I think it could easily be overcome.

I also think accurate vision and positioning could also be achieved, at least to the level possible with a human.

The big question is strategy software. Could the computer ever be programmed well enough with offensive and defensive strategies that could defeat a human in a variety of game situations. IBM developed "Big Blue" to defeat the world champion chess player Kasparov, and strategy is a huge component in chess.

What do you guys think?

I personally think that with enough money, it would be possible to develop a human-conquering pool-playing robot in the next 10-20 years. However, who would sponsor such a project? It would be costly. [Maybe the IPT will be willing to sponsor Dr. Dave and some students. The IPT seem's to have lots of cash.] Also, would it be worth the effort? If the robot beat the best human pro, it would be interesting, but I think it might also be disappointing to some. I, for one, was very disappointed when Kasparov (an extremely smart and creative human) lost to Big Blue (a dumb computer running software, even though the software was written by humans).

I look forward to hearing some of your feedback on this topic.

Regards,
Dave

kyle
12-05-2005, 04:50 PM
I remember seeing this story and thinking that his knowledge of pool was limited, he stated that there's 4 types of spin (right,left,high and low) also the robot looked to emulate his own jacked up stroke. Iron Byron copied a pro golfers swing not the guy who made it.

Billy
12-05-2005, 04:53 PM
<blockquote><font class="small">Quote dr_dave:</font><hr> How long do you guys think it will take for somebody to design a pool-playing robot that can beat the best human pro?

If you watched the videos, one thing that is obvious is that the Canadian robot "breaks like a girl." This is a major liability, but I think it could easily be overcome.

IBM developed "Big Blue" to defeat the world champion chess player Kasparov, and strategy is a huge component in chess.

What do you guys think?

I personally think that with enough money, it would be possible to develop a human-conquering pool-playing robot in the next 10-20 years.

Regards,
Dave <hr /></blockquote>


sorry Dave but you're dreaming if you think a robot could possibly 'play' a sport better than a human even if it didn't 'break like a girl'

moving a chess piece to a particular square could hardly be considered an athletic move even though the software may have dictated where to make the move

and I also don't think you'll be picked for the 'human conquering pool-playing robot' research and experimentation project if ever one.somewhere amongst your numerous posts my scientific evaluations have concluded that it's quite possible some girls may have a better break than of the good dr.

now you may return to your favorite dvd I,Robot

jmo

dr_dave
12-06-2005, 08:03 AM
<blockquote><font class="small">Quote kyle:</font><hr> I remember seeing this story and thinking that his knowledge of pool was limited, he stated that there's 4 types of spin (right,left,high and low) also the robot looked to emulate his own jacked up stroke. Iron Byron copied a pro golfers swing not the guy who made it.<hr /></blockquote>
I agree that the researchers might still have a little to learn about pool, but I cut them some slack in this regard.

I also noticed the jacked up stroke. I guess they don't want to worry about parts of the robot jamming other balls into the table. Humans have much more flexibility with bridge placement, length, and elevation.

Despite all of the negatives, I give those guys lots of credit. What they have done, although not perfect, is pretty damn impressive.

Regards,
Dave

dr_dave
12-06-2005, 08:25 AM
<blockquote><font class="small">Quote Billy:</font><hr>sorry Dave but you're dreaming if you think a robot could possibly 'play' a sport better than a human even if it didn't 'break like a girl'<hr /></blockquote>
I agree that it isn't easy, but I think it is possible. It is certainly possible to design a robot to create a stroke as powerful, versatile, and accurate as that of a human. The tough part is the strategy software. But again, with knowlegeable programmers, and creative software (e.g., that learns from each shot), I still think it is possible. Although, as I wrote, I doubt anyone would be willing to fund such an ambitious project.

<blockquote><font class="small">Quote Billy:</font><hr>moving a chess piece to a particular square could hardly be considered an athletic move even though the software may have dictated where to make the move<hr /></blockquote>
Agreed. Chess is far from being an athletic sport. But pool is much more constrained and athletically limited than many other sports. The table is small and doesn't move, the balls are near perfect spheres and nearly rigid, the playing surface is nearly perfectly flat, the stroke can be a straight-line motion without loss of action, etc.

<blockquote><font class="small">Quote Billy:</font><hr>and I also don't think you'll be picked for the 'human conquering pool-playing robot' research and experimentation project if ever one.<hr /></blockquote>
You're probably right. I already have too many distractions.

<blockquote><font class="small">Quote Billy:</font><hr>somewhere amongst your numerous posts my scientific evaluations have concluded that it's quite possible some girls may have a better break than of the good dr.<hr /></blockquote>
I don't doubt this; although, my break and game have improved a bunch over the last year. I didn't mean to be disrespectful with the "break like a girl" quote. It just seems to be a frequently used phrase lately, with the recent mixing of male and female players in IPT events. Regardless of what comparison one might make (politically correct or not), the robot's break currently leaves a lot to be desired.

Regards,
Dave

Billy_Bob
12-06-2005, 12:13 PM
I think the key (or limiting factor) here is the equipment.

Cloth is different and the *same* cloth can be different depending on humidity/temperature.

Balls are different (friction) and the *same* balls can be different depending on temperature/humidity, how clean the balls are, and new or old and worn.

Cushions can be different (a table can be "long", "short", or just right), and this can change with temperature/humidity.

And how level the table is. I play faster on tables which have a known "lean" so my ball will go straight instead of curving as it would with a very slow shot.

I have my table in my garage and find I can't do certain force follow shots when it is about 45 degrees in there. Yet I can do these when the room warms up or go to a bar with a warm table and my shots work as normal.

So I think the computer would need to read temperature and humidity or have a set of tests to preform on the table first to "calibrate" the system to the table/balls/cloth/cushions first before playing.

And then there is the cue/tip. The same amount of chalk would need to be applied before each shot. Some sort of automatic chalking gizmo.

Then I think the computer experts could figure out the strategy part and vision problems from there.

dr_dave
12-06-2005, 01:19 PM
Billy_Bob,

I think you make some excellent points. I think most people would agree with you that, to be competitive, one (human or robot) must adjust to table conditions.

For the robot to adjust, it would need to calibrate itself by first taking some practice shots (as humans often do). Alternatively, and in addition to this, the robot could update the parameters in its software as conditions change from shot to shot. This is called adaptive learning. The robot could compare the actual shot (from the camera-based vision system) to the trajectory pre-computed by the software, and make adjustments as necessary for future shots. Again, this is the same thing a human must do.

Regards,
Dave

<blockquote><font class="small">Quote Billy_Bob:</font><hr> I think the key (or limiting factor) here is the equipment.

Cloth is different and the *same* cloth can be different depending on humidity/temperature.

Balls are different (friction) and the *same* balls can be different depending on temperature/humidity, how clean the balls are, and new or old and worn.

Cushions can be different (a table can be "long", "short", or just right), and this can change with temperature/humidity.

And how level the table is. I play faster on tables which have a known "lean" so my ball will go straight instead of curving as it would with a very slow shot.

I have my table in my garage and find I can't do certain force follow shots when it is about 45 degrees in there. Yet I can do these when the room warms up or go to a bar with a warm table and my shots work as normal.

So I think the computer would need to read temperature and humidity or have a set of tests to preform on the table first to "calibrate" the system to the table/balls/cloth/cushions first before playing.

And then there is the cue/tip. The same amount of chalk would need to be applied before each shot. Some sort of automatic chalking gizmo.

Then I think the computer experts could figure out the strategy part and vision problems from there.
<hr /></blockquote>

Billy
12-06-2005, 01:41 PM
<blockquote><font class="small">Quote dr_dave:</font><hr>
I don't doubt this; although, my break and game have improved a bunch over the last year. <hr /></blockquote>

then keep on with keepin' on Doc

we need intellectuals who have a passion and appreciate the game itself

jmo

Jal
12-06-2005, 02:50 PM
<blockquote><font class="small">Quote dr_dave:</font><hr>...
It is certainly possible to design a robot to create a stroke as powerful, versatile, and accurate as that of a human. The tough part is the strategy software. But again, with knowlegeable programmers, and creative software (e.g., that learns from each shot), I still think it is possible. ...<hr /></blockquote>

I would think that a robot could surely outdo a human as far as stroke is concerned. People spend a great deal of time trying to make themselves like them in this respect. The robots are already there! And to the extent that the machine can "see" the balls accurately, I would think that they would be better shot makers and execute nearly perfect cueball control. They'll also not be plagued by negative thinking, nervous tremors or self-reproachment.

When you can do all of these things flawlessly (or nearly so) strategy becomes much easier. I'd give Efren the 7,8 and 9.

Jim

dr_dave
12-06-2005, 03:03 PM
<blockquote><font class="small">Quote Jal:</font><hr> <blockquote><font class="small">Quote dr_dave:</font><hr>...
It is certainly possible to design a robot to create a stroke as powerful, versatile, and accurate as that of a human. The tough part is the strategy software. But again, with knowlegeable programmers, and creative software (e.g., that learns from each shot), I still think it is possible. ...<hr /></blockquote>

I would think that a robot could surely outdo a human as far as stroke is concerned. People spend a great deal of time trying to make themselves like them in this respect. The robots are already there! And to the extent that the machine can "see" the balls accurately, I would think that they would be better shot makers and execute nearly perfect cueball control. They'll also not be plagued by negative thinking, nervous tremors or self-reproachment.

When you can do all of these things flawlessly (or nearly so) strategy becomes much easier.<hr /></blockquote>
Jim,

"Easier" is still a relative term. I still think strategy and planning at a top level still might be difficult to teach to a robot with today's computer programming technology. There are so many principles, heuristics, options, and uncertainties that are difficult to put in mathematical and algorithmic terms.


<blockquote><font class="small">Quote Jal:</font><hr>I'd give Efren the 7,8 and 9.<hr /></blockquote>
I'll take you up on this bet. Efren can't lose if you give him the both the 8-ball AND the 9-ball (in a game of 8-ball or 9-ball). /ccboard/images/graemlins/wink.gif

BTW, the Canadian robot is being programmed to play 8-ball.

Regards,
Dave

The_Doctor
12-07-2005, 08:31 AM
Very interesting. What concerns me is how they managed to get funding for what they have already done. I don't know what its like with regards to gaining funding from a university body in Canada but over in the UK I have to submit proposals for my work essentially justifying the greater good that will come from it. Maybe the head of funding at the university is a pool nut. Where are they taking this research and what are they trying to prove with at and then ultimately accomplish? Or maybe they had some spare money left over at the end of the financial year and if they didn't spend it, it would disappear from next years budget. I'm not really suggesting that it is a frivolous piece of work, its just coming from a medical background its difficult to appreciate its worth so if Dr Dave has suggestions.......

dr_dave
12-07-2005, 08:37 AM
<blockquote><font class="small">Quote The_Doctor:</font><hr> Very interesting. What concerns me is how they managed to get funding for what they have already done. I don't know what its like with regards to gaining funding from a university body in Canada but over in the UK I have to submit proposals for my work essentially justifying the greater good that will come from it. Maybe the head of funding at the university is a pool nut. Where are they taking this research and what are they trying to prove with at and then ultimately accomplish? Or maybe they had some spare money left over at the end of the financial year and if they didn't spend it, it would disappear from next years budget. I'm not really suggesting that it is a frivolous piece of work, its just coming from a medical background its difficult to appreciate its worth so if Dr Dave has suggestions....... <hr /></blockquote>

I was also shocked that funding could be found for such a fun project. Maybe the research group can respond to your questions, as I don't know the answers. I talked to the principal investigator yesterday and he said he would try to have someone in the group join in on the discussion.

Regards,
Dave

BigRigTom
12-07-2005, 10:37 AM
My 1st thought is that in the process of making this robot perform there will undoubtedly be new innovations in robotics. Since no one has done it before, those who do accomplish the desired result will learn new ways of solving problems in robotics that will then be used in a multitude of other applications.
Any manufacturing concern that has need of robotics stands to benefit and any computer software concern should also have a reason to pursue this kind of problem solving challenge.
Maybe our friends in Canada should approch MicroSoft, Apple, IBM, Dell, General Motors, Ford, Toyota, Honda and of course there is always Uncle Sam and any other government that is looking to capitalize on knowledge....etc etc.
Being a computer science major, I see it as a possibility and a worthy pursuit and a FUN project idea...I would love to be part of such a project and unlike some of the comments here I know it can be done. It's just a matter of time.
I remember reading about some of the comments that were made regarding Steve Job's vision of a computer in every home. Someone ask <font color="red">"What in the world would anyone want a computer in the home for?"</font color> /ccboard/images/graemlins/confused.gif

dr_dave
12-07-2005, 03:19 PM
<blockquote><font class="small">Quote BigRigTom:</font><hr>My 1st thought is that in the process of making this robot perform there will undoubtedly be new innovations in robotics. Since no one has done it before, those who do accomplish the desired result will learn new ways of solving problems in robotics that will then be used in a multitude of other applications.

Any manufacturing concern that has need of robotics stands to benefit and any computer software concern should also have a reason to pursue this kind of problem solving challenge.<hr /></blockquote>
Their emphasis is on software side of things (vision processing and calibration, game strategy, adaptation, etc.); but I agree with you that it is also an interesting (although not leading edge) robotics hardware project. I hope they get additional funding in the future to also improve the hardware side of things. I want to see a robot break and aim/align/shoot better than any human can. I think it is entirely possible with today's technology.

<blockquote><font class="small">Quote BigRigTom:</font><hr>Maybe our friends in Canada should approch MicroSoft, Apple, IBM, Dell, General Motors, Ford, Toyota, Honda and of course there is always Uncle Sam and any other government that is looking to capitalize on knowledge....etc etc.
Being a computer science major, I see it as a possibility and a worthy pursuit and a FUN project idea...I would love to be part of such a project and unlike some of the comments here I know it can be done. It's just a matter of time.
I remember reading about some of the comments that were made regarding Steve Job's vision of a computer in every home. Someone ask <font color="red">"What in the world would anyone want a computer in the home for?"</font color> /ccboard/images/graemlins/confused.gif<hr /></blockquote>
I also agree that it is an extremely fun project. In fact, I am quite jealous that I don't have such funding. I also agree with you that it would be "only a matter of time" if there is significant and continued funding.

Regards,
Dave

Scott Lee
12-08-2005, 02:08 AM
<blockquote><font class="small">Quote dr_dave:</font><hr>
If you watched the videos, one thing that is obvious is that the Canadian robot "breaks like a girl." This is a major liability, but I think it could easily be overcome.

I look forward to hearing some of your feedback on this topic.

Regards,
Dave <hr /></blockquote>


Dave...Your bourgeois comment is outdated, and sexist. Once again you show ignorance, by promoting the illusion that the only "good" break is a 'power' break...and that women are (apparently, in your opinion) not capable of breaking hard. I still stand by my own professional opinion that anyone, male or female, is capable of breaking accurately and consistently, at 18-20 mph (and can be taught to do so)...which is PLENTY hard enough to spread the balls well, AND control the CB, for either 8-ball or 9-ball. You owe the ladies an apology...

Scott Lee

The_Doctor
12-08-2005, 07:08 AM
Thats a very interesting way to approach research. Thanks very much for replying. [ QUOTE ]


My 1st thought is that in the process of making this robot perform there will undoubtedly be new innovations in robotics. Since no one has done it before, those who do accomplish the desired result will learn new ways of solving problems in robotics that will then be used in a multitude of other applications.
<hr /></blockquote>

Setting ones self a problem, solving it, and then learning things through the journey. A completely different way to other areas of science whereby the question of what effect does x have on y? is asked, this information is then generally used to "suggest" something may possibly be true and therefore have implications for something else. Again thanks for replying I've found that very interesting. Sometimes we all get trapped in our little worlds of comfort and its always good to learn from a completely different disapline.

dr_dave
12-08-2005, 09:13 AM
Scott,

Your points are well taken. I meant no offense. I was just trying to be silly, poking fun at the recent remarks on the IPT threads where the "break like a girl" comment came up. If any ladies (or gentlemen) out there were offended by my remark, they have my sincere apology.

Respectfully,
Dave

PS: Based on the feedback from the people who attended the recent IPT event, it seems that a powerful, accurate break was a distinct advantage. A power break seems to be much less important in 9-ball, where soft breaks can be quite effective. I'm sorry (and take offense) that you think I "showed ignorance" in my posting.

<blockquote><font class="small">Quote Scott Lee:</font><hr> <blockquote><font class="small">Quote dr_dave:</font><hr>
If you watched the videos, one thing that is obvious is that the Canadian robot "breaks like a girl." This is a major liability, but I think it could easily be overcome.

I look forward to hearing some of your feedback on this topic.

Regards,
Dave <hr /></blockquote>


Dave...Your bourgeois comment is outdated, and sexist. Once again you show ignorance, by promoting the illusion that the only "good" break is a 'power' break...and that women are (apparently, in your opinion) not capable of breaking hard. I still stand by my own professional opinion that anyone, male or female, is capable of breaking accurately and consistently, at 18-20 mph (and can be taught to do so)...which is PLENTY hard enough to spread the balls well, AND control the CB, for either 8-ball or 9-ball. You owe the ladies an apology...

Scott Lee <hr /></blockquote>

dr_dave
12-08-2005, 09:24 AM
<blockquote><font class="small">Quote The_Doctor:</font><hr>Where are they taking this research and what are they trying to prove with it and then ultimately accomplish?<hr /></blockquote>
I posed this question to the principal investigator. Here is his response:
<blockquote><font class="small">Quote PI in Canada:</font><hr>
The objective of the Deep Green project is to develop a robotic system that can play competitive pool against a proficient human opponent. We are not just trying to play pool, which is easy: we are trying to play pool *well*, which is extremely challenging for humans and robots alike.

To achieve our goal requires advanced technique and research into four technical areas: mechatronics, computer vision, physics, and computer gaming. Most of our effort has so far been focused on computer vision, although we have made progress in all four areas. As with most new problems, it is first necessary to analyze the problem and acquire an advanced understanding of the various technical challenges
involved before a suitable solution can be achieved. This requires analyzing and understanding the various components of the problem, proposing solution techniques, implementing and evaluating these techniques, and iterating. More often than not, existing solution techniques are not entirely suitable to the new problem, in which case it is necessary to invent new methods. If it were otherwise, then robotic
pool playing systems would already exist!

There have also been a number of other investigations into robotic pool, starting with the Snooker Machine from the University of Bristol in the late 1980s. Since then there have been projects in Japan, Malaysia, Japan, and Iran, as well as the impressive Automatic Pooltrainer from Aalberg
University in Denmark. To put this project in context, in the field of robotics there is a history of setting a challenging research objective and then going about to achieve that goal. An example is the Robocup challenge, which is a large distributed project that aims to build
human-competitve soccer robots by the year 2050. The robotic element of the International Space Station (i.e. the Canadian-made Canadarm) is another example. The Japanese have embarked upon a national incentive to develop walking humanoid robotics which is starting to bear fruit after over a decade of R&amp;D.

So the main justification for pursuing this research is that it is an interesting and challenging technical problem, and we will advance our knowledge by doing so. The elements that are required to solve this problem (computer vision, reasoning and planning, control and actuation) are fundamental to other autonomous robotics systems. If we are ever going to achieve more general and flexible robotic systems, then it is first necessary to succeed with more constrained (albeit currently challenging) problems like robotic pool.
<hr /></blockquote>

Chopstick
12-09-2005, 12:07 PM
<blockquote><font class="small">Quote dr_dave:</font><hr>Where are they taking this research and what are they trying to prove with it and then ultimately accomplish?<hr /></blockquote>

I can see a number of off shoot technologies that this can be applied to. I don't know what bourgeois means, but I'll bet I could build software that would run rings around these guys.

I already know how to fix this things break shot. Think "Rail Gun". /ccboard/images/graemlins/laugh.gif

Doc, if you have a few google minutes sometime check out "emergent behavior". I think that's going to influence a lot of the future direction in AI.

dr_dave
12-09-2005, 12:24 PM
<blockquote><font class="small">Quote Chopstick:</font><hr> <blockquote><font class="small">Quote The_Doctor:</font><hr>Where are they taking this research and what are they trying to prove with it and then ultimately accomplish?<hr /></blockquote><hr /></blockquote>
Chopstick,

This quote was from "The_Doctor," not me. I just wanted to be clear, because in your message, you quoted it from me ("dr_dave"). We are not the same people.

<blockquote><font class="small">Quote Chopstick:</font><hr>I can see a number of off shoot technologies that this can be applied to.<hr /></blockquote>
I agree.

Regards,
Dave

PS: I like your creative "rail gun" idea. That would make for one mean break!

Chopstick
12-09-2005, 01:35 PM
<blockquote><font class="small">Quote dr_dave:</font><hr> <blockquote><font class="small">Quote Chopstick:</font><hr> <blockquote><font class="small">Quote The_Doctor:</font><hr>Where are they taking this research and what are they trying to prove with it and then ultimately accomplish?<hr /></blockquote><hr /></blockquote>
Chopstick,

This quote was from "The_Doctor," not me. I just wanted to be clear, because in your message, you quoted it from me ("dr_dave"). We are not the same people.

<blockquote><font class="small">Quote Chopstick:</font><hr>I can see a number of off shoot technologies that this can be applied to.<hr /></blockquote>
I agree.

Regards,
Dave

PS: I like your creative "rail gun" idea. That would make for one mean break! <hr /></blockquote>

Huh, we got another one in here. /ccboard/images/graemlins/laugh.gif TYPO I was just trying to capture the question. It is an interesting project though. I thought about doing it myself once.

Billy
12-09-2005, 11:59 PM
<blockquote><font class="small">Quote Scott Lee:</font><hr> <blockquote><font class="small">Quote dr_dave:</font><hr>
If you watched the videos, one thing that is obvious is that the Canadian robot "breaks like a girl." This is a major liability, but I think it could easily be overcome.

I look forward to hearing some of your feedback on this topic.

Regards,
Dave <hr /></blockquote>
.

I still stand by my own professional opinion that anyone, male or female, is capable of breaking accurately and consistently, at 18-20 mph (and can be taught to do so)...which is PLENTY hard enough to spread the balls well, AND control the CB, for either 8-ball or 9-ball.

Scott Lee <hr /></blockquote>

Mr. Lee's professional opinion is above

why not say something about pocketing a ball?what good is it to spread the balls well and controlling the cueball if you don't make a ball?this would be highly invited by a player and would clearly say thank you for spreading the balls well and sticking the cue in the middle of the table.why would they say thank you?- because they're now out.it's a roadmap from here Mr Lee

please understand that you do not play profession pool,you only teach professionally to basic and intermediate players.what a tremendous difference there is here

you'll soon have your head handed to you on a platter if you don't spread the balls well AND make a ball

FACT - the harder you break accurately and with control, the more of a chance a ball will eventually find a hole.this is what a player strives for Mr Lee - to make a ball

jmo

HALHOULE
12-10-2005, 09:43 AM
SOFTWARE PLUS TRIG.GO DAVE, GO.

Scott Lee
12-10-2005, 02:43 PM
Billy...I teach all ability players, from beginner to pro! I never claimed to be a "professional tournament player"!
However, I can hold my own with most anyone.

You make some points, but don't know the facts. AccuStats has kept statistics on break percentages for over 20 yrs.
Playing 8-ball, the average professional player comes up dry on the break very near 40% of the time. Some players may have a 'spell' where they do better or worse, during a given match, or perhaps even a given tournament...but the statistics bear out over time. BTW, the percentage is 33% dry on 9-Ball. This is often born out in tv matches. I've heard people say, "So and so made a ball (or more) on the break, 12 games in a row!" Yep, and then that player went to another table, and made nothing for 5 or 6 games. The break is a random shot. You can up your odds, with superb fundamentals, and excellent execution...but consistency is the key. Nobody on earth, even Efren, makes a ball on the break (in any game) all the time...and never will.

Scott Lee
Professional Instructor