Thomas Funkhouser
Papers
21
Total Citations
2,033
H-Index
13
About
Thomas Funkhouser is a prominent computer scientist whose research sits at the intersection of robotics, computer vision, and machine learning, with a particular focus on robotic manipulation, scene understanding, and autonomous systems. His work has fundamentally advanced how robots perceive and interact with unstructured, real-world environments. Among his most influential contributions is his research on robotic pick-and-place systems capable of handling novel objects in cluttered settings without task-specific training data — work that has accumulated nearly 700 citations across its publications. His investigations into dynamic manipulation extended further with TossingBot (284 citations), which demonstrated that robots could learn to throw arbitrary objects accurately by combining learned models with residual physics. His 2017 work on physically-based rendering for indoor scene understanding (277 citations) addressed critical data bottlenecks in training deep neural networks for navigation and assistance tasks. More recently, Funkhouser has explored the integration of large language models into personalized robotic assistance through TidyBot (combined 264 citations), reflecting his forward-looking engagement with emerging AI paradigms. Across his body of work, themes of generalization, data efficiency, and real-world deployability recur consistently, making his research essential reading for anyone pursuing practical, intelligent robotics systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2TossingBot: Learning to Throw Arbitrary Objects With Residual Physics284 citations · 2020
- 3
- 4
- 5TidyBot: personalized robot assistance with large language models189 citations · 2023
- 6
- 7An LSTM Approach to Temporal 3D Object Detection in LiDAR Point Clouds95 citations · 2020
- 8Spatial Action Maps for Mobile Manipulation75 citations · 2020
- 9TidyBot: Personalized Robot Assistance with Large Language Models75 citations · 2023
- 10