Papers
11
Total Citations
172
H-Index
5
About
Masha Itkina is a robotics researcher whose work spans robot learning, embodied AI, and uncertainty-aware perception, with particular emphasis on enabling robots to act reliably in complex, real-world environments. She has made significant contributions to large-scale robot manipulation datasets, most notably through her involvement in DROID, a landmark in-the-wild manipulation dataset that has already garnered over 100 citations since its 2024 release and represents a major step toward generalizable robot policies. Her earlier work tackled environment prediction for autonomous systems, developing attention-augmented ConvLSTM architectures and generative latent occupancy models that help robots anticipate dynamic surroundings under uncertainty. Itkina has also advanced socially aware navigation, proposing occlusion-aware crowd navigation frameworks that leverage human agents as implicit sensors. More recently, she has pushed the frontier of trustworthy robot deployment, investigating failure detection without failure data and statistically rigorous policy evaluation for imitation learning systems — critical challenges as robotic policies grow in complexity. Her research into embodied question answering further demonstrates a broad commitment to building robots that actively reason about and explore their environments. Collectively, her work addresses the full pipeline from perception and prediction to safe, generalizable action, making her a notable emerging voice in modern robotics research.
Research Focus
Key Achievements
Top Papers
- 1DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset108 citations · 2024
- 2
- 3Attention Augmented ConvLSTM for Environment Prediction16 citations · 2021
- 4Occlusion-Aware Crowd Navigation Using People as Sensors12 citations · 2023
- 5
- 6LOPR: Latent Occupancy PRediction using Generative Models4 citations · 2022
- 7DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset3 citations · 2024
- 8
- 9Perception Beyond Sensors Under Uncertainty1 citations · 2021
- 10GHIL-Glue: Hierarchical Control with Filtered Subgoal Images1 citations · 2025