Papers
17
Total Citations
418
H-Index
8
About
Soroush Nasiriany is a robotics researcher whose work sits at the intersection of robot learning, manipulation, and human-robot collaboration. His research addresses some of the most pressing challenges in the field: enabling robots to perform complex, long-horizon manipulation tasks reliably and at scale. Nasiriany has made substantial contributions to imitation learning and reinforcement learning for robot manipulation. His widely cited work on behavior primitives—reflected in papers like "Augmenting Reinforcement Learning with Behavior Primitives" (78 citations) and the PRIME framework—demonstrates how structured abstractions can dramatically improve sample efficiency and task performance. His 2021 benchmark study on offline imitation learning (70 citations) became an important reference for reproducible evaluation in the community. A recurring theme in his research is scalability. Projects like DROID (108 citations), RoboCasa, and MimicGen tackle the data bottleneck in robot learning through large-scale datasets and automated data generation. His human-in-the-loop deployment work further bridges the gap between laboratory demonstrations and real-world robustness. With over 390 cumulative citations and contributions spanning simulation, large-scale data collection, and adaptive deployment, Nasiriany has established himself as a significant voice in making general-purpose robot manipulation both practical and scalable.
Research Focus
Key Achievements
Top Papers
- 1DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset108 citations · 2024
- 2
- 3
- 4Planning with Goal-Conditioned Policies53 citations · 2019
- 5RoboCasa: Large-Scale Simulation of Household Tasks for Generalist Robots27 citations · 2024
- 6
- 7
- 8
- 9
- 10