Papers
5
Total Citations
44
H-Index
3
About
Ben Eisner is a robotics researcher whose work lies at the intersection of computer vision, reinforcement learning, and robotic manipulation. His research focuses on enabling robots to perceive and interact with their environments more intelligently, particularly in complex, real-world scenarios involving transparent objects and continuous control. Eisner’s most impactful contribution is his work on self-supervised transparent liquid segmentation for robotic pouring (20 citations), which tackles the notoriously difficult problem of estimating the state of clear liquids from static RGB images—a critical capability for autonomous cooking and bartending robots. He has also advanced reinforcement learning by developing Q-learning methods for continuous action spaces using cross-entropy guided policies (15 citations), addressing a key challenge in applying off-policy RL to robotics. His notable achievements include the TAX-Pose framework for task-specific cross-pose estimation, which enables robots to manipulate novel objects by learning pose relationships from demonstrations, and a pixels-to-plans approach that combines deep learning with motion planning for non-prehensile manipulation. Eisner’s work consistently bridges perception and action, pushing toward more adaptable, sample-efficient robotic systems.
Research Focus
Key Achievements
Top Papers
- 1Self-supervised Transparent Liquid Segmentation for Robotic Pouring20 citations · 2022
- 2Q-Learning for Continuous Actions with Cross-Entropy Guided Policies15 citations · 2019
- 3TAX-Pose: Task-Specific Cross-Pose Estimation for Robot Manipulation4 citations · 2022
- 4
- 5