Four Labs Get $30M to Build Self-Learning Science Robots

Imagine a robot that gets better at its job every time it works, then teaches other robots what it learned. That's the future four national laboratories are building with $30 million in new federal funding.

The Department of Energy selected four projects to create autonomous robots that can run experiments, learn from experience, and transfer their skills across different labs. Instead of programming robots to repeat the same tasks forever, scientists are building machines that actually improve through practice.

Argonne National Laboratory leads MAESTRO, which focuses on robots that teach themselves. The system uses digital twins and world models to help robots learn from every experiment they run. The goal is simple but revolutionary: create robots that get smarter with use and can adapt their skills to new instruments and environments.

Brookhaven National Laboratory tackles a different challenge with DART. Their project answers a critical question: how do we know when robots are reliable enough for real science? They're building systems to test robot capabilities under stress and verify that the physical results match what experiments require.

Oak Ridge National Laboratory's TRACE project connects everything together. Think of it as creating a universal language that lets robots, instruments, computing systems, and AI agents talk to each other seamlessly. Scientists at one lab could use proven automation from another without rebuilding everything from scratch.

SLAC National Accelerator Laboratory brings it all to cutting-edge facilities with SPIRE. Their robots will handle everything from moving samples to making split-second decisions at electron and photon science facilities, where timing matters down to microseconds.

The Ripple Effect

These four projects work like a team sport. MAESTRO builds robots that learn. DART tests if they're trustworthy. TRACE shares the knowledge across labs. SPIRE puts it all to work on real experiments.

The real win goes beyond individual labs. Every project must create open software, reusable skills, training resources, and datasets that any DOE facility can use. Instead of four impressive demonstrations, scientists are building an infrastructure that the entire research community can tap into.

This approach could speed up discoveries in everything from clean energy to new materials. When robots handle repetitive experimental work and learn what works best, human scientists gain time to focus on creative problem-solving and breakthrough thinking.

The funding starts with $2 million this year, with more coming if Congress approves future budgets. Within a few years, labs across the country could have robots that make scientific discovery faster, smarter, and more efficient than ever before.