Joseph Modayil

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. 1
   Horde: a scalable real-time architecture for learning knowledge from unsupervised sensorimotor interaction

   305 citations · 2011
2. 2
   Factoring the Mapping Problem: Mobile Robot Map-building in the Hybrid Spatial Semantic Hierarchy

   117 citations · 2009
3. 3
   Multi-timescale nexting in a reinforcement learning robot

   68 citations · 2014
4. 4
   Bootstrap learning for object discovery

   60 citations · 2005
5. 5
   The initial development of object knowledge by a learning robot

   59 citations · 2008
6. 6
   Bootstrap learning of foundational representations

   58 citations · 2006
7. 7
   Autonomous development of a grounded object ontology by a learning robot

   47 citations · 2007
8. 8
   Integrating Multiple Representations of Spatial Knowledge for Mapping, Navigation, and Communication

   39 citations · 2007
9. 9
   Building Local Safety Maps for a Wheelchair Robot using Vision and Lasers

   28 citations · 2006
10. 10
    Multi-timescale Nexting in a Reinforcement Learning Robot

    23 citations · 2012

Key Collaborators

Contact & Links