Pure Genius

The physics behind animation

The physics behind animation

Posting in Design

A computer scientist at Columbia simulates motion of everyday objects, with applications in robotic surgery, feature films and the fashion industry.

Photo: Eileen Barroso, Columbia Engineering

During a recent conversation with Eitan Grinspun, I found myself wondering whether he is more obsessed with food or with movement. But then I realized it doesn’t matter, because his detailed explanations of how easily spaghetti breaks or how honey drips are equally—and endlessly—fascinating.

Grinspun is an associate professor of computer science at Columbia Engineering and director of the school’s Computer Graphics Group. He uses computer graphics to simulate the motion of everyday objects, drawing on ideas from graphics, geometry, engineering, physics and, of course, food. But his work also includes making computer-generated hair look real, moving the bristles of a paintbrush and predicting how a material will respond to a collision.

He has developed technologies that have been used in robotic surgery (medical schools), animation of furry creatures (Weta Digital) and princess fashion (Disney). Excerpts of our conversation are below.

Your work combines a lot of areas. Do you consider yourself a computer guy working with physics and geometry or a science guy working with computers?

I think it changes over time. My undergraduate was in engineering science. It’s a program at University of Toronto for bright students who can’t make up their mind about what they want to study. So my background was understanding the world from a physics and math perspective, and by the time I graduated I specialized in computers. Then my masters focused around the design of computer chips. My PhD was in computer graphics.

I realized the more I went into computer science the more I missed geometry, physics and classical mathematics. At the end of the day I still consider myself a computer scientist. One of my professors said what is so great about computer graphics is you can go any direction you want. Which is great if you’re curious. You don’t have to stick to one thing. As long as you’re producing pretty images at the end, people are happy. A lot of our projects have been about something we’re curious about.

The animation process combines art and physics: An animator sculpts the character in an expressive walking pose...

What’s an example?

A couple years ago I was curious about how objects move, especially how flexible material--fabric or sheet metal—moves. I thought about it because it’s a natural progression in geometry. It’s just something I thought about and wanted to do for a while.

I was giving a course at SIGGRAPH (the Association for Computing Machinery's Special Interest Group on Computer Graphics and Interactive Techniques), and at the end of the course Basile Audoly introduced himself and suggested we work on a project together. He showed me the repertoire of things he’d done in the past on the physics and theory side, including the tearing of plastic sheets and breaking spaghetti. He was saying one of the big unsolved problems (he won an Ig Nobel prize for this) is that when you break spaghetti, it always breaks in more than one piece. So he was curious in one side, I was curious in the other side.

In 2007 I ended up flying Basil and a geometer from Germany to Columbia, and we spent a few weeks together looking at making flexible strands move. When we had this working, by total chance, we got a call from Adobe. A new research group said they wanted to pursue a project to develop a physical paint brush that bends and moves. I said, “You’ve got to be kidding me,” because we just, in hindsight, had finished developing a bristle. So that was a successful collaboration. You can see it on Adobe Illustrator 5 and Adobe Paint Brush 5, and you can play with the brush.

...and paints "guide" color patches that instruct the computer in...

What else?

U.C. Berkley and Chapel Hill and Johns Hopkins said they are working on computing how flexible needles can be used by doctors to maneuver around organs without damaging delicate tissues. A needle is just another example of a flexible strand. It wasn’t that we sat down and said, if we do this, we’ll have all these applications at our fingertips. It’s just an abstract shape—a curve is just a shape.

We look at things like honey. If you watch how it falls onto a biscuit and see where it lands, it doesn’t just land and splash like a liquid. It falls like a rope. So why is this liquid behaving like a solid and coiling and folding?

We found experiments that physicists have been doing recently—letting a thick liquid drop on a conveyer belt. What’s amazing is that when it’s falling, it starts to make patterns. If the conveyer belt is moving fast, the liquid makes a straight line, but when it slows down, it coils. So just by varying the speed of the belt you can get more than a dozen very different patterns. They call this the viscous sewing machine.

Other than curiosity about physics, what else drives your projects?

Mathematics and artistic factors. We have a project now being used at Pixar –the idea is that the artists can run a bunch of simulations--maybe it’s a cape or a flag--and in one it looks almost right but you want the cape to strike just the right silhouette. The only way to get these simulations to finish in just a few minutes is to have them be very rough simulations. We call it the butterfly effect—small changes.

The technology we developed, called tracks, or tracking, says you want to track your previous result as much as you can. So that’s an example of a project that’s motivated not by physics but by artistic work.

...automatically introducing wrinkles and folds too tedious to draw by hand...

Tell me about your background.

My parents are from Chile, I was born in Israel, I was brought up speaking Spanish, I learned Hebrew, then we lived in Ann Arbor, Canada and California. Today, I speak Spanish and Hebrew just as fluently as English but I have an accent in all of them. That’s an interesting analogy to the research. Computer scientists will look at me and think I’m more applied physicist; and physicists will look at me and think I’m a more computer scientist.

What’s a typical day for you?

I haven’t had a typical day in a few months because I’ve been having a lot of fun visiting a studio in New Zealand and traveling to conferences.

When I’m at Columbia, the typical day involves teaching a computer animation course to undergraduates or meeting with undergraduates during office hours. The rest of the time I have a bunch of meetings with graduate students where we brainstorm a new direction or they present to me where they’re stuck and I try to work on suggestions with them. We collaborate. I think of them as my peers and I learn from them. That’s easily 80 percent of my day, meeting with them and going over topics. That’s also the hardest part of my job.

My students know that before a meeting I have to make a tea—and part of it is the 10-minite ritual of making the tea and taking the break. I have boxes of loose-leaf tea all around my office. And it’s funny how lab culture evolves and students end up growing around these things. By the time a lot of students gradate, they drink tea.

The last 15 to 20 percent is spent on activities that further the department or the school, like communications or recruiting.

...as well as additional physical effects, such as an electrical shock or a gust of wind. Image credits: Columbia University Computer Graphics Group

What haven’t you animated yet that you’d like to?

This is not a direct answer, but one thing that I always thought would be fun--because I love food--is to do computational cuisine , or computational gastronomy. It could have computers compute the process of having an egg frying, which would depend on the temperature of the oil, the type of the egg, size of the egg, the ambient temperature. It would really capture the ways an egg could cook. An egg is pretty complex, because the yolk cooks at a different temperature than the rest of it. There are shorter-term phenomena that I'm interested in capturing but that’s the thing that hasn’t been done yet.

And among things that sound more serious, one of the things I think is very interesting is something our group has been working on-- the motion of objects that collide with each other. You stomp on a Coke can, and the sheet metal doesn’t go through itself. But what about all the collisions that happen in a full head of hair, or a pile of dirty laundry? So the question is, can we develop computer algorithms that capture the overall behavior of all those interactions without having to compute each single one? Hair has a given bulk because of all the micro-collisions, but can we get that sense of bulk without computing each collisions?

So many questions. They must pop up all the time, in everyday life.

That happens.

Does it keep you up at night?

It wakes me in the middle of the night. When I’m swimming or showering I start to think about these things. Some of the things we weren’t sure about with the flexible strands, I thought about and came up with some answers this way.

Hmm. I also have ideas pop up in the pool or the shower. I wonder if it has something to do with the water.

I just thought the same thing. Maybe it’s because water is meditative.

Share this

Melanie D.G. Kaplan

Contributing Editor

Melanie D.G. Kaplan is a Washington, D.C.- based journalist. She is a regular contributor to The Washington Post and National Parks Magazine. Her website is www.melaniedgkaplan.com. Follow her on Twitter. Disclosure