CORPUS DYNAMICS

The physics of human interaction

We capture how humans physically interact with the real world, so robots can learn to do the same.

Abu Dhabi, UAE

Everyone collects video. No one measures the forces.

When a person picks up a mug, opens a door, or mops a floor, a camera can record every frame of that movement. What it cannot record is how hard they gripped, how their weight shifted, or how much force passed through their wrist. That invisible layer, the measured forces of contact between a human body and the physical world, is exactly what robotics companies need. Startups are racing to collect video of human tasks. No one is collecting the forces.

We deploy lightweight sensor kits (pressure-sensing gloves, smart insoles, and body-worn cameras) through partnerships with facility management and service companies, collecting synchronized video and force data during everyday tasks in real homes, offices, and commercial spaces. The result is a dataset no one else provides: real-world human interaction data with measured contact forces, captured at scale, outside the lab.

ENGINE A

Robotics

Licensed human interaction datasets with real-world force measurements. The training data that robots need and simulation cannot provide

ENGINE B

Healthcare

AI-powered injury risk screening from video, trained on the same real-world force data. No lab visit required

Three layers, stacked

03
The Model

A 989-muscle neuromusculoskeletal model transforms raw sensor data into physics-complete annotations: joint torques, muscle activations, force distributions. The most detailed academic whole-body musculoskeletal model. Cannot be replicated from video.

02
The Data

Pressure-sensing gloves and smart insoles capture measured contact forces, not estimates from video. No startup collects this. Without force data, robots can learn where to reach but not how hard to grip.

01
The Channel

We deploy through facility management companies already operating in thousands of real environments. No per-site negotiation needed. No competitor uses this channel. Every other startup pays gig workers or rents access.

Prof. Yoshihiko Nakamura

Co-Founder & Chief Scientist

  • Dept. Chair, Robotics, MBZUAI
  • Professor Emeritus, University of Tokyo
  • 2025 Japan Academy Prize
  • IEEE Robotics Pioneer Award (2021)
  • IEEE Life Fellow · 32,500+ citations
  • NMS model (989 muscles, 155 DOF)
  • Founder, Kinescopic Inc.
LinkedIn

Kevin Henry

Co-Founder & CEO

  • AI/Robotics Intern, Lockheed Martin CIS
  • Founded Collabrix AI — 2 Mitacs, 1 enterprise customer
  • Winner, UAE MoIAT + EDGE NextGen Challenge
  • Winner, EU-UAE AI for Sustainability Hackathon
  • 2 papers under review (IEEE RA-L, IJCV)
  • Hands-on: Unitree G1, Spot, aerial platforms
  • MSc Computer Vision, MBZUAI
LinkedIn