The Robot Baseball Lab
Joe Liechty
I am a PhD student at the University of Utah. I work with Dr. Tucker Hermans on solving robot manipulation problems using control theory and machine learning.
I am especially interested in how analytical and data-driven methods can be combined to get formal stability guarantees from controllers, do task and motion planning, and create robust controllers in the presence of uncertainty.
I received master’s and bachelor’s degrees in mechanical engineering from the University of Utah and Brigham Young University respectively.
Why Robot Baseball?
The Robot Baseball Lab is not a formal lab (yet), in its current iteration it is a place for me to share all of the things that I’m working on. The idea for the name came from a job interview question I was asked once. The question was, “If you could do anything with [a humanoid robot], what would you do?” My answer was, “I would make it play baseball.” There are a couple reasons for this; 1) it’s cool, and 2) it’s really difficult.
Ted Williams, a former Boston Red Sox’s player and one of the greatest MLB hitters of all time, told the New York Times in 1982 that, “I’ve always said that hitting a baseball is the hardest thing to do in sports. The hardest thing – a round ball, round bat, curves, sliders, knuckleballs, upside down and a ball coming in at 90 miles to 100 miles an hour, it’s a pretty lethal thing.”
Short stop and Hall of Famer Ozzie Smith said that hitting is difficult, “[s]imply because you’re hitting a round object that’s traveling at 95 to 100 miles per hour and you’re trying to hit it on the square part of a round bat…It takes great hand-eye coordination. And the guys who do it and leave this game with an over .300 batting average, I look at those guys as being super, super players.”
A task you only need to achieve 30% of the time in order to be considered successful? Sounds perfect for a robot! All jokes aside though, picking up an object, estimating its kinematics and dynamics, planning how to manipulate it so that it intercepts another object moving at high velocity at a desired position and orientation with an appropriate amount of force, and then executing that plan…just talking about it is a lot! On top of that you need to be able to replan repeatedly as the pitch comes towards you just in case the pitcher was actually throwing you a curve ball or a slider. Doing it all is an incredibly difficult task for a person and an even more difficult for a robot.
A robot that can do this may not be incredibly useful beyond being a cool party trick or a viral YouTube video, but the tools to solve these problems can be applied to other manipulation tasks that are more useful such as surgery, working on a construction project, or working with astronauts on the moon. When the astronaut says, “Hey Robot, pass me that wrench.” you don’t want the wrench to come flying over and smash into their helmet. That is why my personal research projects are focused on this goal: having fun solving the challenging problem of getting a robot to play baseball and develop tools that can be used for other, more practical tasks.