Ellen Novoseller
University of California, Berkeley, DEVCOM Army Research Laboratory
Papers
10
Total Citations
159
H-Index
7
About
Ellen Novoseller is a robotics researcher whose work lies at the intersection of imitation learning, deformable object manipulation, and human-robot interaction. Her major contributions include pioneering interactive imitation learning algorithms like LazyDAgger and ThriftyDAgger, which intelligently manage when and how a human supervisor intervenes during robot training—reducing cognitive burden while maintaining learning efficiency. In the domain of deformable objects, Novoseller has made significant strides in untangling dense, non-planar knots in cables and ropes, developing the IRON-MAN algorithm for multi-cable disentanglement and autonomous strategies for long cables and garment smoothing. Her work on preference-based reinforcement learning further extends robot learning to incorporate human preferences without hand-crafted reward functions. With over 150 citations across her top papers, Novoseller’s research has been recognized for its practical impact on real-world robotic manipulation, particularly in challenging scenarios involving self-occlusion and complex dynamics. Her achievements include advancing sim-to-real transfer for fabric manipulation and developing budget-aware intervention strategies that make interactive learning more feasible for real-world applications.
Research Focus
Key Achievements
Top Papers
- 1LazyDAgger: Reducing Context Switching in Interactive Imitation Learning29 citations · 2021
- 2
- 3Autonomously Untangling Long Cables20 citations · 2022
- 4Efficiently Learning Single-Arm Fling Motions to Smooth Garments20 citations · 2023
- 5Disentangling Dense Multi-Cable Knots19 citations · 2021
- 6
- 7
- 8
- 9LazyDAgger: Reducing Context Switching in Interactive Imitation Learning2 citations · 2021
- 10Disentangling Dense Multi-Cable Knots2 citations · 2021