Independent Physical AI Lab
Physical AI,
field-tested.
EmbodiedEdge builds robots and edge AI systems that hold up outside the demo. Trained in simulation, measured on embedded hardware, and designed for the people who have to run them.
- 01
Simulate
Train and test in simulation first, where failure is cheap and every run can be repeated.
- 02
Deploy to the edge
Run on embedded Jetson compute and measure latency and throughput instead of assuming them.
- 03
Prove it on hardware
Real robots, real networks, and an operator who can review what the system did and why.
Why this lab exists
Most Physical AI never leaves the demo.
A policy that works in simulation is a hypothesis. A system that runs on real hardware, within its power and latency budget, with someone accountable for how it behaves, is a result.
The distance between the two is where most projects stall. The cause is rarely the model. It is integration, measurement, observability, and the operating discipline to keep a system running after launch day.
That gap is the lab's focus. Every project here is meant to end on physical hardware with measured results, and each one says plainly where it stands until then.
Research
Four tracks, one question: does it work on the robot?
Sim-to-real manipulation
Carrying learned policies from Isaac Sim to a physical arm without losing what made them work.
Edge inference
Running perception and detection on Jetson within real power, thermal, and latency limits.
Safety and observability
Instrumenting the whole pipeline so operators see degradation before it becomes a physical failure.
Autonomous navigation
SLAM and navigation on a mobile robot where perception, planning, and control share one small board.
Current builds
What's on the bench now
Physical AI on Jetson
A robot manipulation program that carries a learned policy from Isaac Sim to a real arm, with Jetson inference measured at every step.
Physical AI Safety Observability
A runtime safety layer on Jetson AGX Thor that measures pipeline performance and inference latency, so the system can show when it is operating outside safe limits.
Jetson Edge AI Security
A defensive telemetry runtime for edge devices, pairing live network capture with measured on-device inference and operator-reviewed alerts.
Autonomous Navigation on ROSMASTER
SLAM and navigation on a mobile rover with Jetson Orin NX onboard, starting from measured bring-up instead of assumed performance.
The lab
The lab runs on its own hardware.
Every result on this site comes from equipment in the lab, from simulation on a workstation to inference on the robot itself.
- NVIDIA Jetson AGX ThorPrimary edge target
- NVIDIA Jetson Orin NXOnboard compute, mobile robot
- Synria Alicia-D armManipulation
- Yahboom ROSMASTER M3 ProMobile navigation
- RTX 5090 workstationSimulation and training
Writing
Build logs and research notes
The first build logs are being written. Until they're published, the work is discussed on LinkedIn.
Follow on LinkedInBehind the lab
Obinna Edeh
Obinna brings 25 years of large-scale systems integration leadership to Physical AI, including a nationwide deployment and teams of engineers, field crews, and vendors across multiple markets. EmbodiedEdge applies that operating discipline to robots and edge AI.