A Balancing Act

Drew Warren (MSR ‘18) shares how he built a robot that controls the position of a rolling ball on the surface of a platform

Drew Warren built a balancing platform robot that was able to control the position of a rolling ball.

As Drew Warren considered what type of project to pursue for his Master of Science in Robotics winter quarter independent project, he knew it was important to have a project that kept him interested.

He found that project in the form of a custom Stewart platform.

Warren built a balancing platform robot that was able to control the position of a rolling ball on the surface of a platform. The platform was actuated by six servo motors, and the feedback for where the ball was positioned was provided by a USB camera and specially-developed image processing software.

“This project is a great example of a complete end-to-end robotics problem,” said MSR Associate Director Jarvis Schultz. “There was a custom mechanical and electrical design, complex kinematics and dynamics, computer vision for sensing, simulation tools, control algorithms, and more.”

Warren took time to talk about his project, what he learned from it, and how he thinks the experience will help him moving forward.

What was it about pursuing this project that interested you?

I am particularly interested in learning about control systems, and I like working on real, physical things. I like to solve complicated problems that work in real-life and not just in software or in simulations, and this project allowed me to do that.

In a few sentences, how would you summarize the challenge of this type of project to someone with little understanding of robotics?

It's easy to make a model of a robot or a control system, and it's easy to plug in a motor or a sensor to see how they work. It's hard to put all of these things together on a physical system and make sure that all the pieces work together reliably enough to complete a task. When you integrate all the little pieces of a robot, you learn all the ways that those pieces can fall apart and break each other, and you have to figure out the right way they fit together.

What was the hardest part of the project for you?

The hardest part of this project was making the step from having the thing built and running to getting the ball to stay on the platform. Building the platform, setting up the computer vision, designing the control system — all of that was easy. Getting all of those subsystems to work together enough that the ball was stable and stayed on the platform, ignoring whether the performance was good or not, was really hard.

What were one or two things you learned from this project?

I learned about some basic computer vision tools/techniques like perspective warping, and I learned about linear quadratic regulator (LQR) controllers and how they work.

You were able to present this project at the Museum of Science and Industry (MSI) earlier this year. How did you present it in a way that the general public, including kids, could understand?

First of all, the device worked — they could see the platform continuously moving to keep the ball on the platform. I also made sure I had an interface that was intuitive and simple.  On the screen, I put the video of the image of the platform with an icon on the center of the ball and an icon on the target location. I told them the camera looks to find the position of the ball, compares it with the target position, and then tilts the platform just enough to roll the ball over there but not too much that it falls off. I also put in the work to be able to control that target position with a game controller, so people could move it around and watch as the robot tried to follow along with the ball.

What was the general reaction of the people you spoke with at MSI?

They loved it. I think the fact that they could see this physical device working and that they could interact with it and tell it where to move the ball, I think that's what did it.

How do you think this project will help you professionally moving forward?

Being able to get verified, real-world performance on a physical system is valuable. I have already accepted a job working in the controls department of a defense company, and I think working on this project and others like it prepared me for that.

Learn more about Drew Warren’s project.