Papers
3
Total Citations
133
H-Index
3
About
Jonathan Mugan is a leading researcher in autonomous learning and developmental robotics, with a focus on how agents can bootstrap high-level knowledge from low-level sensory data. His work centers on hierarchical reinforcement learning and qualitative state representation, addressing the fundamental challenge of enabling machines to autonomously discover meaningful actions and states without human-engineered abstractions. His most influential paper, "Autonomous Learning of High-Level States and Actions in Continuous Environments" (2011, 89 citations), introduces methods for agents to learn hierarchical structures using only domain-general knowledge, a critical step toward more adaptable AI. He further advanced this line of research in "Autonomously Learning an Action Hierarchy Using a Learned Qualitative State Representation" (2018, 34 citations), which tackles the problem of learning action hierarchies in continuous domains. His doctoral thesis (2010) laid the groundwork for these contributions, exploring how agents can develop distinctions and actions from pixel-level input. Mugan’s work has been foundational for researchers in developmental robotics and autonomous agents, offering principled approaches to the long-standing problem of scalable, self-directed learning in complex environments.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3