Physical AI for the
world’s most extreme environments.
Ensemble Control is a frontier research lab inventing the sensing and actuation that let autonomous machines perceive and act where humans and conventional robots cannot — in orbit, the deep sea, and disaster zones.
Where we focus our research.
Sensing, actuation, and the learning that ties them together — engineered for environments that break conventional robots.
Sensing
Resilient, multimodal perception and state estimation under radiation, pressure, darkness, and signal loss — so a machine always knows where it is and what surrounds it.
Actuation
Soft, continuum, and high-dexterity actuators that survive and perform where rigid robots fail — like an octopus arm hot-swapping a server rack in orbit.
Physical AI
Sim-to-real reinforcement learning and optimal control that carry policies from millions of simulated trials into the unforgiving real world.
Built for where robots break down.
Our systems are designed for the environments most automation avoids — orbit, the deep sea, and the aftermath of disaster.
Orbital data-center servicing
An eight-arm continuum robot that autonomously hot-swaps server racks on an orbiting data center — trained entirely in simulation, then transferred to hardware.
Open Problems We're Working On
Dr. Abhishek Gupta's talks and lectures on the reinforcement learning, control, and perception that underpin Physical AI for extreme environments.
Reinforcement Learning for General MDPs
TDAI AI Summit, Ohio State University
Perturbation Theory for High-Dimensional Control
General Motors R&D
Reinforcement Learning Course
Ohio State University
Generalized Policy Gradient Theorems
Allerton Conference 2025, UIUC
Collaborate on Research
We work with collaborators across sensing, soft and continuum actuation, and sim-to-real control. If these problems interest you, start a conversation.
Start a ConversationCollaborate with the lab.
Researchers, partners, and prospective collaborators — tell us what you’re working on and how it touches sensing, actuation, or Physical AI.