David Hyde, assistant professor of computer science at Vanderbilt University’s College of Connected Computing, has received a $1.6 million grant from the National Science Foundation to develop a new, open-source software platform for “physical intelligence,” which is the combination of traditional physics simulation algorithms with artificial intelligence for addressing complex real-world applications.
For example, physical intelligence can give robot assistants understanding of physics so they can make better predictions and plans as they navigate the real world. Physical intelligence can also enable safer autonomous vehicles or help industrial designers create more efficient machines and parts. And, as with traditional physics simulations, physical intelligence can also bring physically accurate and visually stunning special effects to the big screen, as in movies like “Star Wars” or “Frozen.”
“Thanks to innovations in AI, GPUs, and the cloud, today’s computing landscape is radically different than ten or fifteen years ago,” said Hyde, the project’s lead principal investigator who is also a member of the Institute for Software Integrated Systems at Vanderbilt. “Yet most of the world’s hardest computing challenges still involve physics. We saw an opportunity to build a new platform, based on these innovations, that combines our expertise in AI and physics simulations into something that researchers and engineers can use to help tackle these challenges.”
The new software platform being built under the grant is called COSTA (a Community Open Simulation, Training, and Applications platform). During the three-year grant period, Hyde and his collaborators will also create a community governance organization, host workshops and hackathons, develop curricular materials, and pursue various other activities to build a thriving base of COSTA users who will use and contribute to the project both during and after the grant period.
Collaborators and Co-PIs include Professor Bo Zhu at Georgia Institute of Technology, Professor Joseph Teran at University of California, Davis, and Professor Ron Fedkiw at Stanford University. The professors, along with their Ph.D. students, will build critical pieces of COSTA and help grow the COSTA developer community.
The multi-institutional team previously collaborated on PhysBAM, a physics simulation library first created by Prof. Fedkiw and used for special effects in countless Hollywood feature films. Hyde, as head of the Simulation, Optimization, and Learning (SOL) Laboratory at Vanderbilt, is currently leading several complementary research projects to COSTA, including in AI and quantum, thanks to a recent NSF CAREER Award.
Funding for COSTA is provided by NSF’s innovative Community Infrastructure for Research in Computer and Information Science and Engineering (CIRC) program. Assistance was also provided by Vanderbilt’s Research Development & Support (RDS).