Letícia Berto’s Post

I will be talking about the "Future of Robotic Assembly" for an ASSEMBLY Magazine webcast next week, combining what we learned in our research in robotic assembly with what is new in machine learning and the emerging field of humanoids. Looking forward to seeing you there!

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

This approach seems to be the most practical for bringing the best parts of AI approaches into mission critical systems. Leverage the unique strengths of AI and classical approaches in combination to mitigate overall system risk and enhance system capability Great work, as always Marco Pavone !

Day 2 of Config, Figma’s excellent design conference, is on tap for today. Day 1 closed with Fei Fei Li’s wide-ranging talk that connected our primordial pre-retina / pre-sight world, the Cambrian explosion, and today’s spatial computing and robotics. “We need AI that can DO.” Fei-Fei is my Shero, and frames today’s advances in spatial intelligence within a human and humane context. #config2024

Imagine a future where robots are so advanced that they have skin that looks, feels, heals, and moves like ours. Would this be super cool or the stuff of nightmares? The work behind this was recently detailed by a team of scientists from the University of Tokyo and Harvard University in Cell Reports Physical Science. They're creating new "perforation-type anchors" inspired by skin ligaments - a unique solution to the skin-attachment problem in robotics. Read more about it here: https://engt.co/3VZwr55

When building new technologies it's important to consider how to make them more inclusive. Here is a recent paper with my colleagues at McGill University that discusses inclusivity in robotics design that will be presented at the 2024 ACM/IEEE international conference on human-robot interaction. https://lnkd.in/eg-e9TJt

In our fresh paper (https://lnkd.in/dhcubuhB) Steve LaValle and I analyze the abstract mathematical theory of robotic déjà vu. If a robot comes to a familiar place it might have seen before, it often needs to realize that it actually has been there before, or, if it hasn't, then realize that this place is new despite looking similar to another place. Failure to do this properly can lead to the robot navigating in circles forever and "thinking" that it is traversing an infinite path. Here by "thinking" we merely mean that the map the robot is constructing in its digital brain is wrong. To mathematically address this problem, we formally define when two robot's environments are equivalent from the robot's perspective (an equivalence relation) and show that the loop closure problem is a consequence of the fact that a space and its covering space are equivalent in this relation. This one-of-a-kind topological treatment of SLAM (simultaneous localization and mapping) is now available as a pre-print on ArXiv and is going to be published by the Eurpean Mathematical Society as a chapter in a book on topological robotics edited by Jesus Gonzalez and Michael Farber.

🚀 Excited to share that our paper "Efficient Parallelized Simulation of Cyber-Physical Systems" has been accepted in Transactions on Machine Learning Research (TMLR)! 🎉 Struggling with the sim2real gap in cyber-physical systems? Our new approach efficiently parallelizes simulations on accelerator hardware, accurately simulating asynchronicity and delays to better mimic real-world interactions. 📈 Collaborative work with Laura Ferranti, Jens Kober, and Robert Babuska. Check out our code example on Google Colab (see link below) demonstrating our supergraph method with Brax to build gym-like environments in #Jax 🔗 paper: https://lnkd.in/eq2kkY3N 📹 video: https://lnkd.in/et4q3Rct 👨💻 code-example: https://lnkd.in/e5N97PsY #CyberPhysicalSystems #ReinforcementLearning #Simulation #MachineLearning #Robotics #AcceleratedPhysics #Sim2real

See you tomorrow! Looking forward to inspiring discussions and questions that will help us at Munich Institute of Robotics and Machine Intelligence (MIRMI) at the Technical University of Munich to understand and #research all perspectives on robotics safety ✨🔬

#AI is a real challenge for lots of businesses. This is a great interview giving some insights into the challenges faced in delivery AI that delivers value. Great to see that #quality is key component part.