Gradually increasing reinforcement learning environment complexity

Question

The problem of robotic arm control has different levels of complexity ranging from simulation to real-world application. For example, a simulation may not model friction of the joints, which becomes important in reality.

Is it possible to use a model trained in a simple environment to bootstrap learning in a more complex environment? Does this have a name? Does it fall under hierarchical reinforcement learning?

Answer 1

This concept is first presented in "Curriculum Learning" by Bengio et at.

From the abstract:

Humans and animals learn much better when the examples are not randomly presented but organized in a meaningful order which illustrates gradually more concepts, and gradually more complex ones. Here, we formalize such training strategies in the context of machine learning, and call them "curriculum learning".