Agility Robotics

Them and Skild AI both are training the motor cortex in the human brain.

Resources

• https://www.agilityrobotics.com/content/training-a-whole-body-control-foundation-model

“This model is a relatively small LSTM neural network with fewer than one million parameters, which is trained in NVIDIA’s Isaac Sim physics simulator for decades of simulated time over three or four days.”.

They talk about 2 sorts of physics:

1. when the legs are actively in the air, moving
2. When it’s making solid contact with the ground, stationary
   • All of manipulation is solved with this assumption

Building a model that works for both is really hard, which is why current robot foundation models only apply for fixed legs.