Intelligence that acts in the physical world.
Machines that sense, decide and act in the physical world are going to be everywhere, on this planet and past it. That is not a forecast in need of defending; the economics are already pushing hard enough to bring it about. The open question is how fast, and it has a physical answer: energy, compute, and the loop between them. The Institute works on that answer, and teaches it, across perception, control and coordination.
Energy-first, physics-informed thinking is how the Institute answers "how fast". Every study here finds the one thing standing between a capability and the field — a material, a price in joules, an assembly, a method, an instrument, a rule, a trained cohort — says what would move it, and says what becomes possible on the other side. Results that went against us are published at the same size as the ones that did not, because a null with the constraint named is a map reference. How a hard thing becomes possible →
Five areas.
The domains the Institute works across.
Perception & World Models
Depth, detection, multimodal fusion, and predictive world models. Perception that holds up outside the lab and runs on the device.
Explore →Manipulation & Dexterity
Contact-rich control, tactile sensing, and bimanual dexterity. Robots that touch, grasp, and assemble to real tolerances.
Explore →Learning & Sim-to-Real
Reinforcement and imitation learning, domain randomization, and embodied foundation models. Policies that survive the real world.
Explore →Digital Twins & Deterministic Physical AI
Reproducible, verifiable embodied AI: a deterministic loop fed by live telemetry and rendered with WebGPU. Includes a live lab.
Explore →Autonomy & Multi-Agent Systems
Coordinated autonomy across air, land, and sea: shared comms, distributed planning, and fleets that hold together under failure.
Explore →The labs.
Where the research happens.
The Glass Lab
The airspace and the people behind autonomous aviation: airspace integration, aviation training, and the STEM workforce.
Enter the lab →The Hiner Lab
Drones in the field: logistics, search and rescue, last-mile delivery, and coastal surveillance, integrated into aviation, maritime, and energy.
Enter the lab →The Hall Lab
What an autonomous system must demonstrate before it is allowed to operate, and who has to be trained to judge it: aviation safety, security, and operations.
Enter the lab →The Charlot Lab
Interface Engineering and Swap-2C Constrained AI: provable, physically grounded AI that runs on the device.
Enter the lab →42 open lines of work.
Each is a standing question with the evidence attached: a technical report where one exists, usually a runnable simulation, and a plain statement of what is still unsettled. Read them as teaching material, because that is what they are written as.
The papers behind them →The lab that runs themThe Control Benchmark standingsFerralloy · the edge-deploy layerFerroscope · the interface layer
Weights you can run.
Our releases ship the weights, the dataset they learned from, and a way to actually run them, on your own device, no server.
nano-world-model
A simulator with no physics engine: a small neural network learned momentum, walls, and the bounce from watching, and now paints every next frame from your action, feeding its own output back. Drive it with the arrow keys.
Play the dream →nano-vla-arm-3d
An open-vocabulary vision-language-action policy running its whole forward pass on-device. Type any instruction, a real language model embeds it on your device, and it drives an articulated arm's joints, from a perspective camera view, to the block you named. 119,986 parameters.
Drive it with words →forge-arm-reach-bc
Our first released policy checkpoint: 17,923 parameters, 72 KB. The exact published weights close the loop live on the same MuJoCo arm they were trained on, at ~100K inferences per second. No cloud.
Run it on the page →describe a rule, not a design
A base case and a recursive rule grow an adder of any width, with no bus size ever declared. The grown graph is lowered to transistors and solved as one nonlinear circuit, so the analog voltages decide whether it computes. Verified from 1 to 16 bits, 480 transistors.
See what grew →certificates and refusal
A safety benchmark counts how often a robot declined an unsafe action. A certificate names the region where refusal is guaranteed, the recover that holds it, and what holding it costs. Three barriers certified as worst cases on an open arm model: the work surface, a person, and the arm's own body.
Read the method →the energy lab
Computing's next limit is electricity, and the reason is not the arithmetic. Watch a calculation pay for itself one operand at a time, then watch a hundred coupled variables fall into an answer without searching for it. Two simulations, the physics behind the substrates being built to escape the memory wall, and a map of eighty-five organisations attempting it.
Run the simulations →physics-fidelity benchmark
The axis appearance hides: does a model obey physics? Eight probes (energy, momentum, contact, friction, identifiability, conservativeness, dissipation and structural prediction) score any dynamics model against analytic ground truth, verified on the real SO-101. A trained black-box world model fails; a structured one, same data, passes. Structure beats data, measured.
See the benchmark →soft bodies, grains & on-device control
The rest of the pure-Rust physical-AI stack, in the page: a volumetric Neo-Hookean jelly wobbling under gravity, discrete-element grains pouring into a settling pile, and a cartpole that swings itself upright by rolling out hundreds of futures in parallel. Every solver verified against a physical invariant, all on the device in front of you.
Run it on the page →the verified browser sensor renderer
Synthetic depth, segmentation, and lidar, the sim-to-real currency, rendered by a pure-Rust signed-distance ray marcher compiled to WebAssembly. Sensor simulation that renders and proves itself, in the browser, on your own hardware, free to anyone, no cloud GPU. The tracked sphere's depth is checked against the exact ray–sphere solution every frame.
Run it on the page →the verified browser fluid bench
A pure-Rust CFD stack, MAC projection and lattice-Boltzmann, compiled to WebAssembly and run in your browser. A lid-driven cavity grades itself against the Ghia (1982) reference table as it converges; a gust bench shows the coefficient model break and the resolved solver reveal the fix; a self-propelled swimmer earns thrust from nothing. Every panel renders the flow and proves itself.
Run the solver →skillpack
The open skill layer: a robot skill is source you own, a policy, a typed contract, a capability manifest, a safety envelope. It gates a skill against your robot's declared capabilities before anything moves, and clamps every action to a runtime envelope. Pick a robot, watch the registry gate skills against it, then corrupt the policy and watch the envelope hold, live in your browser.
Open the live demo →Become an investigator, or intern.
The Physical AI Investigator Program turns an early-career researcher into an independent PI. Students can intern with a lab, and investigators who hold their own funding can bring their program here.