<ul><li>Imitation learning is a popular method for teaching robots new behaviors.</li><li>A new neuro-symbolic imitation learning framework is proposed to teach long, multi-step tasks.</li><li>The framework learns a symbolic representation of the low-level state-action space and uses symbolic planning to generate abstract plans.</li><li>Experimental results show that the neuro-symbolic approach increases data efficiency, improves generalization, and facilitates interpretability.</li></ul>

Neuro-Symbolic Imitation Learning: Discovering Symbolic Abstractions for Skill Learning

Discover more