In the basement of Connecticut’s Beneski Museum of Natural History sit the fossilized footprints of a small, chicken-sized dinosaur. Each track is more than a preserved impression; it’s a data record, holding clues to how the animal once moved. Unlocking that information, however, requires a tool that never existed in the Mesozoic: high performance computing.
Today, advanced simulations can reconstruct those steps in the sand grain by grain, revealing not just how dinosaurs walked, but the core mechanics of locomotion itself.
“We’re looking at using dinosaur tracks, footprints, to reconstruct how dinosaurs moved around 200 million years ago,” explained professor Peter Falkingham, a paleobiologist at Liverpool John Moores University.
Together with postdoctoral researcher Ben Griffin, the pair are combining paleontology, biomechanics, and HPC to peer into the tiniest details of fossilized tracks, down to the millions of sand grains that shaped them. By layering fossil scans with computer simulations, their work replays the exact moment this ancient theropod pressed its foot into prehistoric mud. The implications go far beyond paleontology, offering insights that could influence robotics, medicine, and the science of movement itself.
The Unlikely HPC Pioneers
The scientists behind these simulations are at different stages of their careers, yet share the same goal of using computation to unlock their curiosity. Falkingham is a unique specimen in his own right. With degrees in both paleontology and computer science, he embodies a new generation of researchers who can bridge fossil study with computational power.
“I did my PhD using HPC to simulate dinosaur tracks using finite element analysis,” he recalls. “I’ve always been interested in using computing to explore the biological aspects where paleontology is concerned.”
Falkingham’s current European Research Council grant funds the work of Griffin, a postdoctoral researcher who found himself immersed in HPC not by design, but by chance.
“This was my first experience with HPC,” Griffin explains. “Most of my background was in PowerShell or Windows CMD environments. Switching to Linux command lines was a bit of an adjustment.”
With help from his mentor, he mapped out strategies for running simulations on the LJMU “Prospero” Cluster, which was launched in August 2020. Different nodes handled short test runs or longer, more computationally intensive simulations.
“It’s been pretty painless, for the most part,” he tells us. “Peter very kindly explained how everything worked, and we walked through the setup together.”
Griffin is quick to encourage newcomers to lean on mentorship when learning HPC. “The first thing you want to do is talk to someone who is already using the specific HPC system you want to use,” he advises. “Get a copy of someone’s submission script and have them walk you through it. That will make your life way easier.” He laughs when recalling an early misstep: “We pushed something into the substrate too fast, and particles were exploding all over the place.”
These learning curves highlight an important point: HPC isn’t a gated world reserved for experts. With the right mentorship, accessible systems, and a willingness to experiment, even a “Windows-native” researcher can open the door to large-scale simulation science.
The Need For Speed
To understand why this research demands supercomputing, one must consider the sheer scale of the data being explored. “These simulations are recreating very, very tiny independent particles that represent the grains of the substrate,” Falkingham explains. “If you imagine a chicken-sized dinosaur walking through sand, those grains are about a millimeter across. Even in a relatively small shoebox-sized volume of sand, you’re talking about tens or even hundreds of millions of particles.”
On a standard desktop, that level of computation could take months or even years. “Simulating the interactions between all of those particles requires a massive amount of computational power,” Falkingham emphasizes. “Every particle is interacting, bouncing off the others, and for each of those interactions, you’re calculating forces, accelerations, and other factors. It just doesn’t work on anything less than an HPC system.”
In theory, the team could reduce the particle count by making each grain larger, enough to run on a laptop. But that shortcut destroys the science. “The problem is, once the particles are that large, they start to exceed the size of the toes that are making the footprint,” Falkingham notes. “At that point, you don’t get a realistic footprint anymore.”
And particle count is only part of the challenge. Time and sequence also push the limits. “The need for HPC really comes from the sheer number of calculations that must happen for every tiny fraction of a second in real time,” Falkingham explains.
A single simulation can take days to complete on the Prospero system. Refining the parameters means running multiple full-scale versions. “You can’t just run a quick, small test,” he explains, adding that the physics behave differently at different scales. “You have to go all-in, run a full-scale simulation that might take a day, or even three days, on an HPC system.”
Unexpected Pathways To Innovation
What began as a quest to understand ancient creatures is now sparking collaborations far beyond paleontology. “I promise you, any practical application of this work is purely accidental,” Falkingham jokes. But those “accidents” are opening surprising avenues of exploration.
“We’ve started collaborating with colleagues who have a big robotic arm,” Falkingham shares. This new team is trying to understand how their insights might help robots navigate natural terrain. While legged robots perform well on flat pavement, they often fail on soft, uneven surfaces like sand or mud.
“They’re more interested in how robots interact with deformable substrates and the feedback forces applied when moving through such materials,” he says. By studying how toes spread or curve to adapt, Falkingham notes, engineers can design robotic feet that stabilize machines and extend their range of motion.
The implications don’t stop with robotics. “It’s already well known that walking on sand is helpful for people recovering from hip injuries because sand adapts to imbalanced forces,” Falkingham observes. Their research may someday inform new approaches to physiotherapy or prosthetics. “Can we develop footwear that provides the same kind of passive stability we see in dinosaur and bird feet?” he wonders.
Even sports science is finding applications. The team is collaborating with colleagues who are studying human locomotion, tracking people traversing over sand and mud. “We’re working on simulating those footprints, and then studying the movement back up the limbs,” says Falkingham. These collaborations weren’t part of the original plan, but they underscore how HPC-fueled science often spills over into unexpected fields, creating breakthroughs simply by replaying steps from the past.
Igniting Discovery Across Disciplines
Listening to Falkingham describe the collaborations sparked by this research, it’s clear the impact goes far beyond paleobiology. The work has drawn in robotics, computer visualization, art, and even virtual reality. “A few years ago, we collaborated with a computer visualization team at Brown University on ways to visualize these volumetric tracks. That led to collaborations with our local art school, where students tried to figure out new ways to present the data.”
The scale of the output makes this challenge clear. Each simulation produces hundreds, even thousands, of text files, many with hundreds of millions of lines representing each particle. “What do you do with that?” he asks.
The art students responded in unexpected ways, including using virtual reality to explore motion through time. For Falkingham and Griffin, it underlined the need for communication and interpretation as a key part of the discipline of science.
For Griffin, the process has been eye-opening. “It’s been fascinating to see how this work could apply to other fields, and lead to collaboration across all these different areas,” he reflects. “Yeah, I’d say that curiosity is the main thing that HPC has ignited.”
In their hands, HPC doesn’t just solve problems. It sparks new questions, expanding the boundaries of science and creativity alike.
Connect, Learn, And Innovate At SC25
From capturing the mechanics of a 200 million year old footprint to inspiring robotic design, the story of HPC is one of curiosity sparking collaboration, one discovery at a time. This fall, St Louis becomes the gathering point for that journey as the HPC community convenes for a week of sessions, speakers, and networking.
Join us this November at SC25 to discover how HPC is igniting breakthroughs across science and industry. SC is where disciplines converge: paleobiologists meet computational engineers, visualization experts connect with AI practitioners, and students find mentors who can help shape their careers.
If you want to experience firsthand how the HPC industry is transforming fields, from art to medicine to fundamental science, there is no better place to be than SC25.
Contributed by SC25.
Cristin Merritt is SC25 communications chair and chief marketing officer for Alces Flight Ltd, an HPC consultancy and cloud provider based in the United Kingdom.
Be the first to comment