How to build a better ballplayer

by Skip Derra

So you want to build a better baseball player. How do you proceed?

For most designers, the best building blocks would be the finely honed human traits and instincts copied from the world's best players. How about combining the hitting eye of Alex Rodriguez, the fielding range of Derek Jeter, and the competitive heart of David Ortiz, just for starters?

The story is a bit different for ASU robotics expert Tom Sugar. He can build a better ballplayer by using a camera with a zoom lens and lots more on-board capability.

Sugar is an associate professor of engineering at ASU Polytechnic. He and ASU psychology professor Michael McBeath quite literally are taking a nuts and bolts approach to building a baseball player. Their creation is "Catchbot."

Catchbot is a slick fielding, low-to-the-ground robotic infielder with good range and very good speed. Plus, it exhibits the ball-hawking ability of a Golden Retriever and never asks for star treatment.

The McBeath-Sugar team has developed Catchbot for seven years through several generations of robots. The robot ballplayer employs a number of basic scientific principles of perception, movement, and interception. The ASU researchers say their work could lead to advances in robotics. In addition, Catchbot is helping to explain why birds, bees, and even humans are so good at catching.

In its current version, Catchbot is part robot and part miniature dune buggy. It is equipped with two motors, a suspension system, four-wheel steering, and four-wheel drive transmission. The robot has an on-board camera for eyes. It can move up to 30 feet per second.

Catchbot uses a bumper for a "mitt." The resolution of its camera is 640 x 480 pixels and 30 frames per second. With this level of sophistication, Catchbot has exhibited a fielding percentage of .750 for grounders. Not bad, according to its developers, but there is plenty of room for improvement.

Sugar says that more on-board computing will give the robot the capability to sense an image and analyze it on its own. In its current form, the machine must send the image back to a laptop computer for analysis. The result is then sent back to Catchbot before action can occur.

The ASU robotics expert says that a better camera will give Catchbot greater range. Right now it can only see an object up to 20 feet away.

"Sometimes we forget how good the human eye really is," McBeath says, referring to Catchbot's visual limitations.

The research has other more utilitarian applications. For example, Sugar says, a robot that can field a ground ball also can be used to retrieve objects for a disabled person or to help a blind person intentionally avoid a collision.

Besides, intercepting objects is as old as animals hunting on Earth. Extending that to the study of fielding grounders is just a cool application of perception and robot research.

Read more about Catchbot in "Can the robot come out to play?"