Research scientists spend their careers seeking that “Aha!” moment when experiment becomes discovery. Most experiments fail, sending researchers back to the drawing board. But what if the experimental cycle—hypothesis, experiment, data, interpretation, next experiment—moved a hundred times faster?
It’s happening now at Pacific Northwest National Laboratory (PNNL) through autonomous science. We are building continuously learning scientific platforms that connect AI, robotics, and data architecture to power through frustrating failures and get to “Aha!” moments in days instead of months or years. Our goal is for autonomy to touch every experiment in five years.
Autonomy begins with automation, but it doesn’t end there. In truly autonomous systems, data pipelines, decision-making AI, and robotics operate as one tightly coupled and validated system.
It can sense when conditions drift from the desired range, adopt a new strategy, and adjust to emerging information, while documenting each adjustment to maintain rigor. The hardware thinks. The software acts. And the whole system learns.
Across PNNL, in fields as diverse as biotechnology and critical materials separation, autonomous science is already yielding solutions to thorny research questions. Early successes are building momentum across our campus as autonomy is built into new mission areas, from chemistry to biology to new energy storage applications.
Autonomy in the lab
A proving ground for autonomy: the Autonomy Studio
The Autonomy Studio provides a purpose-built space for running design-build-test-learn cycles using digital twins—with their physical counterparts sitting right next to them.
Inside, researchers can prototype and connect the building blocks of an autonomous laboratory, including robotic platforms, sensors, and control systems, then measure performance in realistic workflows. A digital twin extends the studio beyond its walls, allowing teams across PNNL to model how autonomous systems could operate in their own lab spaces before hardware is deployed. The space is outfitted with motion capture to strengthen the feedback loop between simulation and reality, improving precision, coordination, and repeatability as systems learn.
Prototype an experiment in simulation and then test it on-site. Fix the glitches and run it again. It’s a proving ground to perfect autonomous protocols and robotic handlers before moving into an operational lab.
Autonomy in battery materials innovation
Meet MIRAL—the Materials Innovation through Robotics & AI Laboratory. It was designed to analyze thousands of experimental organic molecules and liquid electrolytes for redox flow batteries, lithium-ion batteries, sodium-ion batteries, and many other battery chemistries for energy storage. It helps researchers characterize and predict material structure, solubility, electrical conductivity, and performance. MIRAL’s intelligent database makes it easier to organize and understand complex data related to the chemical properties of materials.
Part robot, part workstation, part intelligent database. MIRAL doesn’t assist researchers. It runs the experiment: designing, executing, and learning from hundreds of trials a day. What used to take weeks now takes hours. The result: hundreds of experiments per day, with fewer errors than manual workflows—because every run is guided by physics-informed datasets that sharpen the next one.
Autonomy in the field
Maritime sensing and autonomy
At PNNL-Sequim, autonomy leaves the lab and hits the open ocean. This is where we rapidly mature technology from the bench to the bay.
An autonomous surface vehicle persistently monitors marine systems for longer, over wider areas, and with more consistency: no fatigue, and no gaps in the data. Pair it with aerial drones, and you get multi-platform teaming: coordinated robots covering complex environments from the surface and the sky.
Autonomous observation platforms on commercial ships
Boundary Layer Exploration of Aerosols and Clouds ON Ships (BEACONS) demonstrates autonomous deployment of a shipborne observing system that routinely measures atmospheric, cloud, and aerosol properties without requiring a dedicated research vessel.
By delivering consistent, AI-ready observations at scale, BEACONS helps scientists resolve aerosol-cloud interactions and supports the Department of Energy mission to improve predictive weather models.
