It's Friday afternoon, and we’re packing for L4DC 2025 @l4dc_conf next week 😎 Excited to share MS-HGNN — our take on bringing morphological symmetry into GNNs for legged robot dynamics. 🧵 (1/4) We bake both kinematics and morphological symmetry into the model → -

Catch us at Board 4.16 during Poster Session 4 (4:45–5:45 PM) in the Pendleton Room, Michigan Union — we’d love to chat or just say hi! Looking forward to seeing you there 👋 #L4DC2025

🧵 (4/4) And how does it actually perform? 🔹 Contact detection (Mini-Cheetah, real) → F1 ≈ 0.9 with just 5% of training data → Uses only 38% as many parameters as ECNN 🔹 GRF estimation (A1, sim) → Smoother 1D & 3D force predictions on unseen terrains → Beats

🧵 (3/4) How do we actually build in symmetry? We made MS-HGNN symmetry-equivariant by: 1. Decomposing robots into kinematic branches 2. Applying group actions (C₂, K₄, etc.) 3. Using Cayley graphs to structure connections 4. Wrapping it all with symmetry-aware encoders &

🧵 (2/4) So… why care about morphological symmetry? Most legged robots are symmetric by design — Go2 (C₂), Mini-Cheetah (K₄)... and symmetry isn’t just pretty — it’s powerful 💪 We started from MI-HGNN, which captured kinematic structure. With MS-HGNN, we go further — now

Lunar Lab @ #ICRA2025 @GTrobotics @GTaerospace

🚀 Gearing up for ICRA 2025 in Atlanta! Our Lunar Lab @ Georgia Tech is bringing the heat: 🌟 MI-HGNN & MS-HGNN for legged robot dynamics 🤖 STATE-NAV for humanoid traversability and navigation 🎙️ Prof. Lu Gan @ganlumomo speaks at two workshops: Foundation Models and

Chances are good you'll spot a bot 🤖 if you're near the Georgia World Congress Center in Atlanta next week. These are some of the types of robots and their skills you can expect from @GeorgiaTech. #ICRA2025 Learn more 🔗: https://t.co/vAOdP5m7fj

📣 Rankings are out! For the second year in a row, #GTaerospace ranks No. 2️⃣ among graduate aerospace engineering programs in the nation, No. 1️⃣ for publicly-funded institutions, according to @USNews and World Report. 🚀: https://t.co/b7EYKhvrNh

We're thrilled to host a lab tour and engaging robot demos for local K-12 students during National Robotics Week! Excited to inspire the next generation and share the amazing world of robotics. 🚀🤖 #RoboticsWeek #STEM #Outreach @GTrobotics @GTaerospace

Today at 12:15 = Multi-Modal and Multi-Robot Coordination in Challenging Environments with  Sebastian Scherer from the Robotics Institute at Carnegie Mellon🤖🎉   @CMU_Robotics @GT_EPIC_Lab @gregory_sawicki @gtcomputing @gatechengineers #autonomy #Robotics #swarm #coordination

📣 Call for Contributions to the #ICRA2025 Doctoral Consortium (DC)! We invite PhD students working in the broad research areas in robotics and automation to apply. 🔗Learn more about the submission requirements here:

📣 We invite contributions to the #ICRA2025 Late Breaking Results Poster Session! Present your latest findings in the field of robotics and automation. 📆 Submit a poster in PDF format by April 11. 🔗 Learn more about the submission details here:

.@Robojackets Join us at 12:15 in Klaus for Computational Symmetry and Learning for Roboticsnwith Maani Ghaffari from @UMRobotics Come grab a lunch and enjoy it with some great listening. Full Deets Here:  https://t.co/9szndbn2Ur] @GhaffariMaani    @GTResearch @gvmcmurray

🚀 Rising Star Workshops for Junior/Senior PhDs, and Postdocs! 🌟 Don't miss these career-boosting opportunities! https://t.co/CbFcjfnyom Please share with your peers, students, and anyone who might benefit! #PhD #Postdoc #Academia #RisingStars

🤖🚨 I'm looking for a PhD student (School of Electrical Engineering and Computer Science at the University of Queensland, Brisbane, Australia 🦘) to work on an exciting that could lead to a paradigm shift in robotic spatial perception and state estimation! The project is at the

We are co-organizing the ICRA'25 Workshop on Foundation Models and Neuro-Symbolic AI for Robotics. See you in Atlanta!

(4/4) Although MI-HGNN is applied to contact perception problems for quadruped robots in this work, it can seamlessly extend to other multi-body dynamical systems and has the potential to improve other robot learning frameworks.

(3/4) We evaluate the proposed MI-HGNN on two common contact perception tasks: contact state detection (classification) and ground reaction force estimation (regression). The former is evaluated on a real-world Mini Cheetah dataset collected by @JustinTzuYuan , while the latter

(2/4) The architecture and connectivity of the MI-HGNN are derived from the robot's morphology (kinematic structure). By incorporating morphology-informed constraints into the neural network, we improve the learning-based approach using model-based prior knowledge. Based on