Pool-Playing Robot: Deep Green

13:27 Thu 17 Sep 2009
[, , , , ]

The Robotics and Computer Vision Lab at Queen’s University Ontario has produced a working robotic system that can play pool. Called Deep Green, it appears to have an excellent understanding of geometry, although it’s not clear that it understands the rules of the game per se, or that it can do its own shot selection.

TechCrunch cover Deep Green here, and include the following video:

In it, they also demonstrate “augmented reality pool“, a system that determines the angles being shot and projects them onto the pool table. This is presumably a projection of the system that Deep Green itself uses to figure out how to aim, like a rather more sophisticated version of “The Spider”, a pool training aid.

I suspect that projects like these, despite seeming frivolous on the surface, generate significant progress in robotics and artificial intelligence. Whatever about the future, the current state of AI means that its successes will come in limited domains, and games are excellent testbeds for this.

As for pool specifically, it’s possible that robotic pool players will be a luxury accessory at some point in the near future. Meanwhile, the assisted-vision aspect of it may well develop from requiring a setup worth tens of thousands of dollars, with perfect lighting and multiple cameras, projectors, and computers, to requiring merely a contact lens overlay and maybe a small separate camera—potentially bringing to pool halls everywhere the joys of aimbot use accusations.

Leave a Reply