Digital Discovery #RSCDigital’s Post

Good eg of biomimicry

🦿 New research out of Georgia Tech explores a unified control framework for robotic lower-limb #exoskeletons. 🌉 This work aims to bridge the gap between in-lab exo technology and real-world human ambulation. 🔎 Learn more in the Science Robotics publication linked below.

Researchers Wanxin Jin and Michael Posa from Arizona State University and Modular Robotics Lab (GRASP Lab) illustrate how to use just 4 minutes of experiential data to learn a model for robust real-time manipulation of a previously unknown object in a recent T-RO paper. https://lnkd.in/emsKCgHp Paper title: Task-Driven Hybrid Model Reduction for Dexterous Manipulation Authors: Wanxin Jin and Michael Posa Paper link: https://lnkd.in/emsKCgHp Image caption: One rollout by running the learned reduced-order g -MPC on the three-finger manipulation system for the Cube Moving task. (b) shows the snapshots of the environment at key time steps of the g -MPC rollout. Here, the upper row of (b) shows the whole environment, and the lower row show the zoom-in details. Explanations and analysis are given in Section VII-D2 and Table VIII. #DexterousManipulation #ModelPredictiveControl #MPC #HybridModel #Robots

One thing that's lacking in the current designs of "Conversational AI" is... Conversation! Ha! Human conversation is a joint activity between partners, and the activity includes turn-taking and interactive alignment. In other words, 1. You know when to say after your conversation partner 2. You are constantly adjusting what and how you say to speak more similarly to your partner's speech Without these features, human conversation would be so inefficient and annoying. We all know a person or two who is like that, unfortunately. And, currently, those "Conversational AI" systems can often be such annoying conversation partners. Have you experienced that? I have, a lot. Recently, two research presentations caught my eye: 1. Regarding turn-taking, Professor Gabriel Skantze's research: https://lnkd.in/gtwkJPf5 2. Regarding interactive alignment (or entrainment), Professor Roger Moore's research: https://lnkd.in/gsx5Hyn9 I was really glad to know that efforts are being made to incorporate these essential features of human conversation into human-computer interaction devices. I am also planning to contribute my own work to this field, so I hope to update you sometime soon.

🚨Tomorrow!🚨 Join Professor Gerry Lacey, Dept of Electronic Engineering, for a webinar on the effect of Robotics and AI in Agriculture. With an overview of the MSc in Robotics and Embedded #AI 🤖 To register: https://shorturl.at/fxGO5

Finally, very excited to share about our paper, TTGO, which has been accepted in the International Journal of Robotics Research (IJRR)! Out of all of my papers during PhD, this is the one that I am most satisfied with, even though the main idea did not come from me and I was not even the first author! The excellent idea was proposed by Suhan Shetty 😀 You can find more info about the paper on our website: https://lnkd.in/dPmPZ_mT

🚀 Thrilled 🚀 to share that our paper with Teguh Santoso Lembono , Tobias Löw and Sylvain Calinon titled "Tensor Train for Global Optimization Problems in Robotics" has been published in the International Journal of Robotics Research! 🤖📚 Key Highlights: • Introducing TTGO, a novel approach for optimization of functions in Tensor Train (TT) representation. • TT, a powerful class of Tensor Networks, serves as an expressive function approximation tool that can efficiently model many functions commonly encountered in robotics (e.g., cost functions for various robotics tasks) • Demonstrated robustness on benchmark functions for numerical optimization and real-world robotics challenges like inverse kinematics with obstacles and motion planning. • Bonus: Handles both continuous and discrete variables and can provide multiple solutions, making it an invaluable asset for policy learning in robotics. For more info: https://lnkd.in/dpzUGF-E Grateful for the collaborative efforts that made this work possible! 🙌 Looking forward to the impact in #Robotics #ML #RL #Optimization 🌐💡

It's an honor to finally see our paper published in one of the most impactful robotics journals, the IEEE Transactions on Robotics! Our paper introduces a human-inspired scene perception model for mobile service robots, leveraging neuroscience concepts, which have been surveyed in our previous paper (https://lnkd.in/eyZTwCiC).  The approach integrates fundamental neuroscience concepts, dividing perception into recognition, knowledge representation, and knowledge interpretation. The recognition system separates background and foreground to incorporate image-based object detectors and SLAM. A multi-layer knowledge base represents scene information hierarchically, offering interfaces for high-level control, while knowledge interpretation methods deploy spatio-temporal scene analysis and perceptual learning for self-adjustment. We evaluated the model's components through a fetch-and-carry scenario in two simulated and one real-world environment, using a single-setting ablation study. --> Improvements for Navigation: The approach enhances performance through the overlap of heatmaps with the robot's costmap, helping the robot to avoid crowded areas or search within these areas for objects of interest. --> Improvements for Manipulation: By splitting the background and foreground, the perception provides pre-processed, semantically enhanced regions, enabling efficient motion planning. Link to the full paper (open-access): https://lnkd.in/emmQ85Kq Thanks to my co-authors Jochen Lindermayr, Dr. Birgit Graf, Werner Kraus and Marco Huber !

Implantation of a total artificial heart offers a solution for patients with severe heart failure, but existing artificial hearts have major limitations, which means there is a need for a better alternative. Through his doctoral research, Luuk van Laake has contributed to the development of a future artificial heart based on soft robotics. https://lnkd.in/gfDPVtmU

Check out all of the webinars from LRIG-NYC's "Automate and Innovate: Robotics in the Lab" series on our YouTube channel!