This makes the task harder, since the neural network must now learn to generalize to more randomized environments. As the neural network gets better at the task and reaches a performance threshold, the amount of domain randomization is increased automatically. This frees us from having an accurate model of the real world, and enables the transfer of neural networks learned in simulation to be applied to the real world.ĪDR starts with a single, nonrandomized environment, wherein a neural network learns to solve Rubik’s Cube. To overcome this, we developed a new method called Automatic Domain Randomization (ADR), which endlessly generates progressively more difficult environments in simulation. Factors like friction, elasticity and dynamics are incredibly difficult to measure and model for objects as complex as Rubik’s Cubes or robotic hands and we found that domain randomization alone is not enough. The biggest challenge we faced was to create environments in simulation diverse enough to capture the physics of the real world.
