Running with the Dinosaurs
Gearing Up with a Mac Cluster
Welding a frame to transport a multi-ton section of Dakota’s mummy, protectively coated for a trip to the NASA CT scanner in Canoga Park, California. (photo courtesy National Geographic)
What made Dakota run, presumably, is Tyrannosaurus rex, the renowned predator of the Cretaceous. The University of Manchester team has determined that Dakota, despite his girth, could easily outrun T.rex.
“Dakota could really shift,” says Manning. “T.rex could do about 18 MPH. But hadrosaur dinosaurs like Dakota could do around 25 MPH. As my colleague Professor Bill Sellers says, ‘T.rex was running for lunch. Dakota was running for his life.”
Dakota has gone to Bismark, ND for the rest of the restoration process, but his CT scan data is fueling research into his lifestyle. To simulate Dakota’s walk, Manning and Sellers, Lecturer in Integrative Vertebrate Biology at the University of Manchester, laser-scanned bones from another hadrosaur skeleton, imported them into Maya, a 3D modeling application, then scaled them up using measurements from Dakota’s block (all done on a Mac Book Pro). After three months they had worked out a multitude of physiological details, including the length of muscle fibers and tendons and how they attached to the skeleton. Then they faced the challenge of simulating Dakota’s walk – with no living animals that can serve as a reference.
“Nothing on the planet today walks like a dinosaur,” says Manning. “And when we put our work into the computer for the first time and ran a simulation, it looked like the Monty Python Ministry of Silly Walks.” In fact, it takes thousands of iterations to get a hadrosaur’s walk right. Sellers acquired a cluster of eight Mac Pro systems to do the job.
“I’ve used Macs for years and years,” says Sellers, “and I wanted a local cluster. With a big remote computer, you spend all your time copying files back and forth. With my cluster, I can just sit on the main server and do all the graphics work at the same time. It saves an enormous amount of time. I’m not a systems admin by any means, so I’m glad these machines just sit there running quietly. They seem to manage quite happily.”
Sellers uses Xgrid to run his Mac Pro cluster, and has just upgraded to Leopard (Mac OS 10.5), which has Xgrid upgrades that simplify his work. He invites colleagues across the campus to contribute unused compute cycles to his project; Xgrid lets him know whether their workstations qualify to join the compute grid.
Doing the Hadrosaur Walk
Sellers uses the cluster’s 60 cores (seven 8-core systems and one quad-core) to run his own GAITSIM software, developed in Xcode, to simulate musculoskeletal movement. GAITSIM is easily parallelized, so each core runs a simulation. Sellers uses a genetic algorithm to calculate the gait pattern, then applies ODE, an open-source physics engine, to calculate things such as joint constraints and the way the feet interact with the ground. The goal is to find the muscle activation pattern that produces a physiologically realistic gait.
“With my cluster, I can just sit on the main server and do all the graphics work at the same time. It saves an enormous amount of time”
In the first 1000 simulations, the hadrosaur may not manage a step without falling on its face. But the software selects the motion pattern that lets it lurch the furthest ahead, and runs another 1000 variations based on this modest success. The iterative approach eventually yields an animal that can walk, then run.
“I don’t constrain the method of locomotion,” says Sellers. “I just give the simulations 5 or 10 seconds in half or quarter real time, to move as far as they can any way they can. So occasionally we get things like skipping dinosaurs.”
How about those animated dinosaurs in Jurassic Park?
“That’s animation,” says Sellers. “In the real world, they’d fall over.”
Professors Phillip Manning and Bill Sellers. Their work on a Mac Pro cluster is bringing Dakota back to life.
Sellers uses one of the Mac Pros as a client server to manage the cluster and watches progress on a 23-inch Apple Cinema Display. When the simulation needs to be dressed up for a presentation, Sellers uses POV-Ray, an open-source ray-tracing system, to put flesh on the model.
Beyond Dakota
Finding out how Dakota ran is an intermediate goal for the project. The knowledge acquired can help scientists to visualize the ecosystem he lived in, his energy budget, his diet, and other elements of his ancient world. The biomechanics and locomotion studies will help biologists better understand how people and animals move today. It will aid in the design of prostheses and robots. This may not be the primary reason for Seller’s involvement, however.
“There are practical uses for this knowledge, but some of us never grew up,” he says. “We’re still fascinated by dinosaurs. And this is a good way to interest kids in science. They think we’re all boring people who stand around in white coats.”
Manning speaks up in the same way for paleontologists everywhere. “I go to conferences on math, physics, and computer engineering. I work with materials scientists, biologists, and computer engineers. Paleontology isn’t about a bunch of guys at the bottom of a dirty hole digging up bones. It’s dino CSI.”