Jake Welde

TL;DR: join us at ICRA 2026 for debates and current work on the future of geometry in robotics! 🤖 https://lnkd.in/eBR9VeSa ICRA 2026 Workshop: "Geometry in the Age of Data-Driven Robotics" If you’re attending ICRA 2026, don’t miss this full-day workshop (on Friday, June 5 📅) exploring one of the most timely questions in robotics today: "What is the role of geometric methods 🌐 in an era increasingly dominated by data-driven approaches 📊?" Geometry 📐 has long been a cornerstone of robotics, shaping how we model, plan, and control robotic systems. At the same time, learning-based methods 🧠 are rapidly transforming the field. This workshop brings these perspectives together 🤝, sometimes in harmony, sometimes in tension, to critically examine where geometry stands, where it struggles, and how it may evolve. Instead of a schedule dominated by invited talks, this workshop emphasizes discussion, debate, and community engagement 💬: - Three rounds of structured pro/con panel debates with leading experts  - Active audience participation, including live Q&A and interactive polling  - Final open panel discussion to synthesize insights from the day  - Live streaming via Zoom, with shared interaction tools (e.g., Slido, PollEverywhere) so remote participants can actively engage In the debates, panelists will argue for 👍 or against 👎 the following claims 💭: 1. Education 🎓 :  “The mathematical foundations of geometry are a top priority for robotics education today.”  2. Research 🔬 : “The vast majority of interesting research questions on the role of geometry in robotics have already been answered.” 3. Communication 📝: “Writing papers and giving talks using formal mathematical jargon only serves to increase the field’s barrier to entry.” We are excited to hear the perspectives of expert panelists 👥 from around the world 🌍: - Noémie Jaquier, KTH Royal Institute of Technology - Park Jong Woo (Frank), Seoul National University - Andreas Müller, Johannes Kepler University - Stefano Stramigioli, University of Twente - Patrick Wensing, University of Notre Dame - Ross Hatton, Oregon State University - Søren Hauberg, Technical University of Denmark - Georgia Chalvatzaki, Technical University of Darmstadt - Zac Manchester, Massachusetts Institute of Technology - Bruno Vilhena Adorno, University of Manchester - seth hutchinson, Northeastern University - Nadia Figueroa, University of Pennsylvania We will also solicit contributions (📨 submission deadline: April 15) from early-career and junior researchers to present ongoing work related to: - Geometric methods in robotics - Learning-based alternatives to classical geometry - Hybrid model-based and data-driven approaches Submit your contributions here ✍️: https://lnkd.in/e7PVvWfH Organizers: - Riddhiman Laha, Northeastern University - Tobias Löw, University of Washington - Jake Welde, Cornell University ——

If you’re at IEEE CDC in Rio 🇧🇷, catch me today at 10:00am talking about flatness-preserving residual learning! TL;DR: Unmodeled Effects 💨: “That’s a nice differentially flat system you’ve got there. Would be a shame if some residual dynamics broke its flatness!” Us 🚁: “not so fast - try learning Flat Residuals!” This work was led by Fengjun Yang and in collaboration with Nikolai Matni.

Thanks Cornell Mechanical & Aerospace Engineering for putting together this lovely profile! I'm so glad to be here. My research group - the Geometry, Design, and Control Laboratory ("GeoDesiC Lab") - is actively recruiting students at all levels! We're looking for control theorists, electromechanical designers, hardware hackers, and more. Find out more about the lab here: https://lnkd.in/ehVK-pYA 🤖 🚀 🛰️ 🌐

A bit late posting, but I had a fantastic time co-organizing the Equivariant Systems workshop with Robert Mahony, Stephan Weiss, and Maani Ghaffari at this year's RSS Conference in Los Angeles. So thrilled with the fantastic contributions from workshop participants! If you're interested in symmetry in robotics, you can find recordings of all talks (including the contributed lightning talks) on the workshop website: https://lnkd.in/etUubqmq (Unfortunately, the panels did not record due to room audio issues.)

It’s with great excitement that I share some big news: after defending my doctoral thesis this summer, I will join the faculty of the Sibley School of Mechanical and Aerospace Engineering (Cornell Mechanical & Aerospace Engineering) at Cornell Engineering! I couldn’t be more thrilled to soon become part of this one-of-a-kind academic community as I launch my research program and begin teaching my own courses. Before heading to Ithaca, I’ll spend the next couple months wrapping up nearly a decade here at Penn Engineering and the GRASP Lab. Most of all, I am deeply grateful to the long list of mentors, colleagues, family, and friends whose help along the way got me to this point.

Looking forward to organizing this exciting workshop at RSS with outstanding colleagues! If you're excited about symmetry-informed methods in robotics, this is the workshop for you, whether you're new to the space or want to share your latest work! Think of this workshop as the frequently-requested "gentle introduction" to Lie groups and equivariance, along with the chance to discuss cutting-edge results!

At IEEE CDC in Milano 🇮🇹? You can catch me twice this week: 1. At the “Meet the Faculty Candidates” Poster Session (yes, I’m on the job market! 📬) on Monday 6-7pm, presenting “Geometric Abstractions for Efficient and Explainable Control of Complex Robotic Systems”, my vision for how geometry and control can play vital roles in bringing robots 🤖 closer to the incredible complexity and variety of Nature 🦅 🐍 🐠 🦍 🐿️. 2. In the "Aerospace" Session on Tuesday 11:20-11:40am, presenting “Almost Global Asymptotic Trajectory Tracking for Fully-Actuated Mechanical Systems on Homogeneous Riemannian Manifolds”, joint work with Vijay Kumar that extends almost global asymptotic tracking control 💫 for mechanical systems from Lie groups to the more general class of homogeneous Riemannian manifolds 🌐. Such methods have applications in reduced-attitude tracking for axisymmetric satellites 🛰️ or as the building blocks of hierarchical controllers for underactuated systems 🚀.

Our work (with Nishanth Rao, Pratik Kunapuli, Dinesh Jayaraman, and Vijay Kumar) just won “Best Contribution - Neuroscience & Interpretability” (one of three best paper awards) at the Workshop on Symmetry and Geometry in Neural Representations (NeurReps) at NeurIPS 2024 in Vancouver 🇨🇦! We prove that Markov decision processes describing tracking control problems 💫 in robotics 🤖 exhibit natural symmetries that can be used to construct a corresponding quotient MDP of lower dimension. Our numerical experiments empirically demonstrate that this reduction 📉 improves sample efficiency and test-time performance 📈 when training tracking controllers via reinforcement learning. “Leveraging Symmetry to Accelerate Learning of Trajectory Tracking Controllers for Free-Flying Robotic Systems”, Jake Welde*, Nishanth Rao*, Pratik Kunapuli*, Dinesh Jayaraman, and Vijay Kumar (*equal contribution) 🎥 Video: https://lnkd.in/eBvA8vsH 📄 arXiv: https://lnkd.in/etA-rUpR 🌐 Website: https://lnkd.in/e84pmJdG 💾 Code: https://lnkd.in/euWC8tjf P.S. if you’re at IEEE CDC in Milano 🇮🇹 this week, come hear about this work during the “Meet the Faculty Candidates” Poster Session on Monday 6-7pm!