DrEureka uses open-source software to train robots, like a robot dog balancing on a yoga ball, by simulating tasks in a 'sim-to-reality' system before real-life execution.
Domain randomization in DrEureka randomizes factors like wetness and mass in simulation, enabling the robot to learn and adapt virtually through a reward/punishment system based on AI-generated code.
[
add
]
[
|
|
...
]