Robot Learning: Generalization In Planning Across Morphologies

In my last lecture on large-scale #robotlearning, I discussed one of my most interesting directions, generalization across sequences and robots. This generalization across sequences of actions or states is a challenging, data-intensive process that requires experience across various robots and tasks. In this lecture, I cover concepts related to learning representations that enable plans to generalize. I make connections between system identification, sim-to-real, meta-reinforcement learning, and in-context learning, and discuss the limitations and generality of each method.