Papers
9
Total Citations
205
H-Index
5
About
Anusha Nagabandi is a robotics and machine learning researcher whose work sits at the intersection of model-based reinforcement learning, robot control, and dynamics modeling. She is best known for her pioneering contributions to sample-efficient deep reinforcement learning, particularly her highly cited 2018 paper "Neural Network Dynamics for Model-Based Deep Reinforcement Learning with Model-Free Fine-Tuning" (68 citations), which demonstrated how learned dynamics models could dramatically reduce the data requirements of robotic skill acquisition compared to purely model-free approaches. Nagabandi has made significant strides in dexterous manipulation, with her 2019 work on deep dynamics models for multi-fingered robotic hands (67 citations) tackling some of the most challenging fine motor control problems in robotics. Her research also extends to the miniaturized frontier, developing image-conditioned neural network controllers for underactuated legged millirobots—platforms valued for their mobility and low cost. More recently, she has explored meta-reinforcement learning from visual observations through latent state models. Spanning human-robot trust, multi-robot localization, and advanced manipulation, Nagabandi's body of work reflects a versatile and impactful research trajectory that has meaningfully advanced the field of data-efficient robot learning.
Research Focus
Key Achievements
Top Papers
- 1
- 2Deep Dynamics Models for Learning Dexterous Manipulation67 citations · 2019
- 3
- 4
- 5Cooperative inchworm localization with a low cost team12 citations · 2017
- 6Trust During Robot-Assisted Navigation5 citations · 2013
- 7
- 8
- 9MELD: Meta-Reinforcement Learning from Images via Latent State Models3 citations · 2020