Vibhavari Dasagi

About

Vibhavari Dasagi is a robotics researcher specializing in reinforcement learning, robot manipulation, and the integration of classical control methods with modern machine learning techniques. Her work addresses some of the most pressing challenges in applied robotics: sample efficiency, safe learning, and real-world deployment. Among her most recognized contributions is her 2020 work on fabric folding (23 citations), demonstrating that robots can learn to manipulate deformable objects — a notoriously difficult problem — through just one hour of self-supervised real-world experience. This breakthrough highlighted the potential of data-efficient learning for contact-rich manipulation tasks. Dasagi's influential Bayesian Controller Fusion framework (21 citations) elegantly bridges traditional hand-crafted controllers with deep reinforcement learning, offering a hybrid strategy that is both sample-efficient and practically deployable. Complementing this, her Ctrl-Z work (13 citations) tackles training instability in reinforcement learning, proposing principled recovery mechanisms critical for safety-sensitive robotics applications. Her broader research portfolio explores sim-to-real transfer, residual and multiplicative controller fusion for navigation, and task-agnostic exploration strategies — collectively advancing the field's ability to move learned robot behaviors from simulation into unpredictable real-world environments. With over 80 total citations, Dasagi's work represents a thoughtful and impactful contribution to practical, deployable robot learning.

Research Focus

Key Achievements

Top Papers

1. 1
   Learning Arbitrary-Goal Fabric Folding with One Hour of Real Robot\n Experience

   23 citations · 2020
2. 2
   Bayesian controller fusion: Leveraging control priors in deep reinforcement learning for robotics

   21 citations · 2023
3. 3
   Ctrl-Z: Recovering from Instability in Reinforcement Learning

   13 citations · 2019
4. 4
   Bayesian Controller Fusion: Leveraging Control Priors in Deep Reinforcement Learning for Robotics

   7 citations · 2021
5. 5
   Residual Reactive Navigation: Combining Classical and Learned Navigation Strategies For Deployment in Unknown Environments

   6 citations · 2020
6. 6
   Multiplicative Controller Fusion: A Hybrid Navigation Strategy For Deployment in Unknown Environments

   6 citations · 2020
7. 7
   Multiplicative Controller Fusion: Leveraging Algorithmic Priors for Sample-efficient Reinforcement Learning and Safe Sim-To-Real Transfer

   3 citations · 2020
8. 8
   Zero-shot Sim-to-Real Transfer with Modular Priors.

   2 citations · 2018
9. 9
   Evaluating task-agnostic exploration for fixed-batch learning of\n arbitrary future tasks

   2 citations · 2019

Key Collaborators

Contact & Links