Papers
16
Total Citations
456
H-Index
7
About
Homer Walke is a leading researcher in scalable robot learning, whose work has fundamentally advanced how robots acquire generalizable skills across diverse platforms and environments. His primary contributions center on developing large-scale datasets and open-source policies that enable robots to learn from heterogeneous data rather than task-specific training. Walke is a core contributor to the landmark Open X-Embodiment project, which produced the RT-X models and datasets—cited over 220 times collectively—demonstrating that training on diverse robot embodiments yields robust, transferable manipulation policies. He also led the creation of DROID, a large-scale in-the-wild manipulation dataset (108 citations), and BridgeData V2, which provides over 60,000 trajectories for scalable robot learning. His work on Octo, an open-source generalist robot policy, and SuSIE, which leverages pretrained image-editing diffusion models for zero-shot manipulation, further exemplifies his impact. Walke’s research consistently shows that robots can achieve broad generalization without starting from scratch, making him a pivotal figure in the movement toward foundation models for robotics.
Research Focus
Key Achievements
Top Papers
- 1
- 2DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset108 citations · 2024
- 3Open X-Embodiment: Robotic Learning Datasets and RT-X Models101 citations · 2023
- 4Octo: An Open-Source Generalist Robot Policy66 citations · 2024
- 5BridgeData V2: A Dataset for Robot Learning at Scale12 citations · 2023
- 6
- 7Octo: An Open-Source Generalist Robot Policy8 citations · 2024
- 8
- 9
- 10