Research
Why they fall over.
Watch any humanoid combat demo closely and the same thing happens. Contact is made and both machines go down. That failure is not a detail to edit around. It is the actual frontier, and it is what we work on.
Four open problems.
01
Push recovery.
A robot takes a force it did not predict and has to stay standing. Humans do this with hundreds of small stabilising muscles and instincts we never think about. Machines have torque and a control loop. Generating force events in simulation and training a model to generalise its recovery is live research with real published work behind it.
02
The sensor to action gap.
There is no autonomous robot fighting today because the distance between sensor input and a decision is enormous. When do you throw, when do you block, when do you wait. Every sport is built on reading what your opponent is about to do, and machines currently cannot.
03
Motion interpolation from human capture.
Put a person in VR, capture the motion, and get a robot to follow it. This mostly works until forces enter the picture. Your body balances through a kick with feedback the VR rig never captured. Closing that interpolation gap so the result is not funky is a genuine problem.
04
Sensor dropout and uncontrolled contact.
Real machines lose sensors mid action and take contact from angles nobody simulated. Behaving sensibly when the inputs degrade is unglamorous and it is what separates a demo from a system.
Simulation first, then the floor.
The stack is a physics capable simulation environment for generating force events and training recovery behaviour, and then real hardware in a lab in Miami to find out which parts of that survive contact with the world.
The event is not separate from this. A card full of machines taking real contact in front of an audience is a source of failure cases you cannot get in a lab, and the parts that break in public are the parts worth working on next.
What comes out of it.
Published work.
Papers out of the lab and out of the companies in the incubator.
Better machines.
Capability improvements that show up on the floor at the next event.
Talent.
Research is how you become the place serious robotics people in the Southeast want to be.
We are early. There is no publication list yet, and when there is, it goes on this site whether or not the results were flattering.