Heni Ben Amor
@asurobot
Visiting Faculty @ Google DeepMind | Associate Professor for Robotics and Machine Learning at Arizona State University. Director of Interactive Robotics Lab.
I am a Googler! 😀 Happy to say that I joined @GoogleDeepMind as a Visiting Faculty Researcher for my sabbatical. I am extremely delighted to work with some of the smartest and kindest people on the planet. If you are also in the Bay Area and want to grab coffee, let me know :)
Self-improvement for robotics @GoogleDeepMind: In this article, I describe together with @pannag_ recent research on robot self-improvement, LLM-based training, and self-play. We had fun trying this on a table tennis robot. Article: spectrum.ieee.org/deepmind-table… #Deepmind #ASU
Robot table-tennis from @DeepMind and @ASU 🤖🏓: We present a new method that allows Gemini (LLMs) to teach and coach robots through self-improvement. The method called SAS Prompt will be presented at ICRA 2025. More info here: sites.google.com/asu.edu/sas-ll… #Deepmind #ASU #Robots #LLM
We’re helping robots self-improve with the power of LLMs. 🤖 Introducing the Summarize, Analyze, Synthesize (SAS) prompt, which analyzes how they perform tasks based on previous actions and then suggests ways for them to get better using the medium of table tennis. 🏓
Simon is one the smartest people I ever met! He is on the faculty market🥳 He wrote the super-influential NeurIPS paper that started the "Language-Conditioned Robot Learning" revolution [shorturl.at/cZhgd] (180 citations). Let's see who wins the competition to hire him 🤖
📣 Thrilled to announce that I'm on the job market for Fall 2025 faculty positions! I am currently a postdoc @CarnegieMellon @CMU_Robotics. 🔍 My research is dedicated to developing robots that can intelligently reason about their environments and the humans within them. By…
I really like this video from our recent article on robot self-improvement @GoogleDeepMind . We (with @pannag_) describe how self-play, LLM-based training and self-improvement can be used to scale-up robot learning. Check it out: spectrum.ieee.org/deepmind-table… #Deepmind #ASU
We are presenting our paper on LLMs for robot self-improvement for robot table-tennis tomorrow at ICRA 2025! Looking forward to see you all. #Deepmind #ASU
4/11 3️⃣ SAS-Prompt: Large Language Models As Numerical Optimizers for Robot Self-Improvement Session: Representation Learning 4 When: Wed May 21 | 16:45 - 16:50 Where: Room 311 | WeET8.3 (collab w/ @asurobot and group)
This week I gave an invited lecture at @mavrojean's Computational HRI class. Chris is doing amazing work at the University of Michigan. Very grateful for the invitation and fun interactions with the students :)
It was an honor to have @asurobot join for a guest lecture on ML for HRI in my Computational HRI class. Fantastic talk, kept students engaged all along. Many thanks! fluentrobotics.com/comphri/2024
Latest work from my lab "Stateful Diffusion Policies" - a combination of deep state-space models and diffusion. The result isa powerful policy representation for Lon-horizon and multimodal policies. #ASU #IROS2024
As we keep exploring the scalability of diffusion-based policies, it’s always important to make the robot history-aware. Check out our new #IROS2024 paper on improving diffusion-based policy by adding a transition model (ControlNet). @asurobot @yfzhoucs diff-control.github.io
The one and only @JeffDean talking about the table tennis robot at @GoogleDeepMind. Humbled to have worked in this project. :) @pannag_ @lgraesser @samindaa @atiliscen @Evo_Daveo @AlexBewleyAI @leilatakayama
Really excited to see us talk publicly about the progress on this work. A few people in our awesome robotics team many years ago came to me and suggested that we should work on this problem (even though it would require buying some pricey higher speed robots) because it would…
We @GoogleDeepMind simply couldn't have achieved this breakthrough without the powerful #MuJoCo simulation engine. Its flexibility, ease of use, and advanced modeling capabilities were absolutely critical to our research velocity. 🚀 #AI #Robotics #TableTennis
Our robot first trains in a simulated environment, which can model the physics of table tennis matches accurately. Once deployed to the real world, it collects data on its performance against humans to refine its skills back in simulation - creating a continuous feedback loop.