Papers
19
Total Citations
886
H-Index
12
About
Joseph Modayil is a prominent AI and robotics researcher whose work spans reinforcement learning, autonomous robot knowledge acquisition, and spatial mapping. He is perhaps best known for his foundational contributions to the **Horde architecture** (2011, 305 citations), a scalable framework enabling robots to learn rich world knowledge through unsupervised sensorimotor interaction using parallel reinforcement learning sub-agents called "demons." This influential work laid important groundwork for modern approaches to continual and predictive learning in AI systems. Modayil has also made significant contributions to developmental robotics, demonstrating how learning agents can autonomously bootstrap object ontologies directly from raw sensory experience — without human supervision — a capability central to truly autonomous intelligence. His research on "nexting" (2014, 68 citations) further explored how robots can continuously generate short-term predictions about their environment, mirroring cognitive mechanisms observed in biological agents. In mobile robotics, his work on hybrid spatial-semantic mapping (2009, 117 citations) combined topological and metric representations, inspired by models of human cognition, to enable robust robot navigation and communication. Across more than a decade of research, Modayil's cumulative contributions have helped define how intelligent systems can learn grounded, structured knowledge from experience rather than explicit programming.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3Multi-timescale nexting in a reinforcement learning robot68 citations · 2014
- 4Bootstrap learning for object discovery60 citations · 2005
- 5The initial development of object knowledge by a learning robot59 citations · 2008
- 6Bootstrap learning of foundational representations58 citations · 2006
- 7Autonomous development of a grounded object ontology by a learning robot47 citations · 2007
- 8
- 9Building Local Safety Maps for a Wheelchair Robot using Vision and Lasers28 citations · 2006
- 10Multi-timescale Nexting in a Reinforcement Learning Robot23 citations · 2012