Vahe Taamazyan’s Post

Had a fantastic time at the IEEE Humanoids conference in Nancy, France! 🇫🇷 It was an honor to present our work on humanoid multi-contact locomotion in constrained environments. I am grateful for the opportunity to connect with brilliant researchers and for the many insightful talks and conversations throughout the conference. Looking forward to building on this momentum and pushing the boundaries of humanoid robotics! Link to the paper below, for those interested. 𝗧𝗶𝘁𝗹𝗲: Guiding Collision-Free Humanoid Multi-Contact Locomotion using Convex Kinematic Relaxations and Dynamic Optimization 𝗣𝗿𝗲𝗽𝗿𝗶𝗻𝘁: https://lnkd.in/gY6KJXer

I am attending this week the International Symposium of Robotics Research (ISRR), which is taking place in Long Beach, CA and is celebrating 40 years since the first event. I had the opportunity to present a survey/position paper on "The State of Robot Motion Generation", which my students Joe Doerr, Patrick Meng, Sumanth Tangirala and I wrote with the objective of reviewing the broad spectrum of techniques that have been developed for robot motion generation across many sub-disciplines. The paper concludes with a set of recommendations for how we can potentially benefit from integrating ideas from different methodologies. Paper link: https://lnkd.in/gaV6-uGX Joe Doerr: https://lnkd.in/gxEHXiPk Patrick Meng: https://lnkd.in/ggP_SQnQ Sumanth Tangirala: https://lnkd.in/gb3dy4p9 ISRR 2024: https://lnkd.in/gq9fdPwS

I am excited to share our latest blog post titled "Genesis: A Generative and Universal Physics Engine for Robotics and Beyond." This article delves into the innovative capabilities of Genesis, highlighting how it serves as a game-changing resource for advancing robotics and other applications. Discover how this transformative engine not only enhances the efficiency of robotic systems but also paves the way for groundbreaking developments across various fields. Read the full post and explore the potential of Genesis here: [Genesis: A Generative and Universal Physics Engine for Robotics and Beyond](https://ift.tt/0WUfNLH)

🚀⚛️ The Summit of Teamwork: New World Record by Mitsubishi Electric Engineers! 🔥 Rubik's Cube Solving Record in 0.3 SECONDS! 🧮 The algorithm used to solve the Rubik's Cube involves moving the layers in a structured order: steps such as cross-forming, solving corners, edge alignment and orienting the final layer, ensuring guaranteed solution. 👉🏼 Click “Follow” not to miss the developments in robotics https://lnkd.in/dfe7gmwH

Last weekend, I had the opportunity to attend my first international conference, IEEE RAS HUMANOIDS Conference. In this work, we proposed a strategy to de-weight outliers measurements that can compromise the accuracy of the state estimation module in a quadruped robot. This paper was a collaboration between the LabRob at Pontifícia Universidade Católica do Rio de Janeiro and DLS at Istituto Italiano di Tecnologia. Not only did I have the chance to present our paper, but I also got to explore the city of Nancy and immerse myself in cutting-edge advancements in robotics. Here are some special moments for me: ► The interactive poster session (If you're curious to learn more, here’s the full paper https://lnkd.in/dqfpx3vk) ► Get an up-close look at some amazing robots in action: in the video we can observe a variety of robots, including a Wall-E-inspired robot from The Walt Disney Company, a humanoid robot from PAL Robotics navigating an obstacle avoidance competition, a quadruped robot from Ghost Robotics and a friendly robot from Enchanted Tools ► Nancy is also the birthplace of the mathematician Henry Poincaré, I end up finding a street named after him I would also like to thanks to my advisor João Carlos Virgolino Soares and my friends Angelo Bratta and Lorenzo Amatucci for the support.

I’m giving a talk at ICRA this afternoon in the Annex Rm 206 at 4:30pm. Given that real world experiments are difficult to carry out at scale, how can we have better confidence in the pre-trained visual representations (PVRs) that we develop for robotics? We carried out an empirical study to see how predictive evaluation in simulation is to real robot performance. Building off of the work we did in making the PVR VisualCortex-1 (https://eai-vc.github.io/) and curating CortexBench (17 tasks meant to measure the generalizability of a visual representation), our study covered 5 PVRs and 5 tasks. More details here: https://lnkd.in/eNTd5iWU

At RSJ2024 (The 42nd Annual Conference of the Robotics Society of Japan), I presented my research on a method for generating robot arm behaviors through imitation learning based on bilateral control utilizing visual and haptic information. The Q&A and discussions after the presentation were very enjoyable! I have refined the ACT learning model employed in systems such as ALOHA, and am currently implementing it within a framework of position and force control. I invite you to kindly visit the English version of our project page to explore related research. ↓our method https://lnkd.in/dhRjD5dM

At this year's IROS, we'll run a workshop on "Robot Safety in Times of AI: Data, Decision, and Multimodal Interaction" covering topics from safe HRI and safety verification to safe decision-making in robotic systems. If you are interested in submitting a paper: we'll be happy to receive _your_ paper, the deadline (20th of September) is approaching, and we have prizes for the best three contributions! Munich Institute of Robotics and Machine Intelligence (MIRMI) at the Technical University of Munich | IROS 2024

Happy to share that the final journal paper from my PhD was published in Soft Robotics! We use embedded layer jamming to dynamically change the stiffness of sections along a growing vine robot. By doing so, we can selectively “activate” discrete joints and control many degrees of freedom with just a few actuators. This also enables new actuation patterns not possible without stiffening. Read the manuscript here: https://lnkd.in/ezQV5Aye Big thanks to my co-authors Shuai Wu, Ruike Renee Zhao, and Allison Okamura for making this work possible.