Constraining policies to select behavior modules allows reinforcement learning transfer from simulation to real robot swarms.

In the spirit of my recent “Math 2.0” talk on new ways to use AI assistance to enhance the human understanding of mathematics ...