Leon Keller

Papers

Total Citations

H-Index

About

Leon Keller is a pioneering roboticist whose work sits at the intersection of neuro-symbolic AI and open-ended learning. His research focuses on solving two fundamental challenges in robot learning: enabling machines to master long-horizon, multistep tasks, and equipping them with the ability to autonomously discover diverse, reusable skills. Keller’s most influential work, "Neuro-Symbolic Imitation Learning," proposes a novel framework that learns not just isolated actions but the symbolic abstractions underlying complex behaviors—a critical step toward robots that can reason about their own learning. Complementing this, his earlier paper "Model-Based Quality-Diversity Search" introduces a powerful algorithm that combines quality-diversity optimization with learned forward models, allowing robots to efficiently explore and build rich behavioral repertoires without exhaustive real-world trial-and-error. While both papers currently hold 2 citations, they represent a forward-looking synthesis of symbolic reasoning and data-driven learning. Keller’s work is particularly notable for its ambition to bridge the gap between short-skill imitation and the compositional, abstract understanding required for truly autonomous robots, positioning him as a rising voice in the next generation of robot learning research.