- Ngoc Nguyen
- 0 Comments
- 1052 Views
📌 Abstract: In this talk, Dr. Quan will discuss the historical development of the robotic manipulation team at Google and recent efforts to scale robot learning at Google DeepMind. He will begin by describing our model scaling effort, starting with robotics-transformer-1 [1], a scalable architecture for learning low level robotic actions. He will then introduce robotics-transformer-2 [2], their next generation model built on top of vision language models demonstrating the transfer of web knowledge directly to low-level control. Afterwards, he will present their data scaling efforts, robotics-transformer-x [3], which demonstrates a foundation model for low level robotic control that exhibits positive transfer across datasets from different robotic embodiments. robotics-transformer-x is a collaboration with 34 research labs across 22 different institutions across the world. He will also discuss their effort in leveraging the code writing ability of language models to enable zero-shot generalization in robotic manipulation [4]. To end the talk, he will discuss their efforts in going beyond high quality demonstrations, q-transformer [5] for leveraging suboptimal data, and going beyond language as a conditioning modality, including using trajectory [6] and sketches [7] to convey the intended tasks.
[1] https://robotics-transformer1.github.io/
[2] https://robotics-transformer2.github.io/
[3] https://robotics-transformer-x.github…
[4] https://openreview.net/forum?id=OOx6a…
[5] https://qtransformer.github.io/
[6] https://rt-trajectory.github.io/
[7] https://rt-sketch.github.io/
📌 About our speaker: Dr. Quan Vuong (Researcher at Google DeepMind) Quan Vuong is a researcher in the robotic manipulation team at Google DeepMind. He obtained his PhD from the University of California San Diego with Dr. Henrik Christensen and Dr. Hao Su. He recently led the robotics-transformer-x project, which is a collaboration with 176 researchers across 34 different research labs to build large scale datasets and generalist foundation models for robotics. His works were recently featured on the New York Times [8], TechCrunch [9] and various other news outlets.