Papers
15
Total Citations
1,898
H-Index
10
About
Roozbeh Mottaghi is a leading figure in Embodied AI, a field that trains intelligent agents to perceive, navigate, and act within physical environments. His research bridges computer vision, robotics, and deep reinforcement learning, with a focus on enabling robots to generalize across new tasks and spaces. Mottaghi’s seminal 2017 paper, "Target-driven visual navigation in indoor scenes using deep reinforcement learning," with over 1,500 citations, tackled two critical bottlenecks in the field: the inability of agents to generalize to new goals and their severe data inefficiency. This work laid the groundwork for more practical, real-world navigation systems. He further advanced the field by co-creating the AllenAct framework (2020), a standardized platform that has become a cornerstone for Embodied AI research. More recently, his GOAT system (2024) enables mobile robots to navigate to any semantically specified object over extended periods, while his work on Habitat 3.0 pioneers the simulation of collaborative human-robot tasks. As a key contributor to the HomeRobot project, Mottaghi is also pushing the frontier of open-vocabulary mobile manipulation, aiming for robots that can understand and act on natural language commands in unstructured homes.
Research Focus
Key Achievements
Top Papers
- 1Target-driven visual navigation in indoor scenes using deep reinforcement learning1,507 citations · 2017
- 2
- 3AllenAct: A Framework for Embodied AI Research44 citations · 2020
- 4GOAT: GO to Any Thing37 citations · 2024
- 5
- 6Navigating to Objects Specified by Images23 citations · 2023
- 7
- 8Habitat 3.0: A Co-Habitat for Humans, Avatars and Robots15 citations · 2023
- 9
- 10HomeRobot: Open-Vocabulary Mobile Manipulation13 citations · 2023