BOME! Bilevel Optimization Made Easy: A Simple First-Order Approach (arxiv.org/pdf/2209.08709…) Bo Liu(Bo Liu ), Mao Ye, Stephen Wright, Peter Stone(Peter Stone ), Qiang Liu We, Nov 30, 16:30 -- Poster Session 4

Synthesize LPV state-feedback controllers directly from data? While guaranteeing stability and performance of the closed-loop? With only a single sequence of data? Wut?! Check out our new paper "Direct Data-Driven State-Feedback Control of LPV Systems" arxiv.org/abs/2211.17182

Course announcement: "Computational Control" bsaver.io/teaching Next spring Saverio Bolognani will bring MPC, RL, DeePC, etc. into the ETH Zurich bachelor curriculum!

Recommendations from Yann LeCun talk at SSL #NeurIPS2022 Workshop

CoRL was really fun this year! My notes for Conference on Robot Learning are available here: cs.brown.edu/people/er35/no…

This guy right here never sleeps. Happy to be back from my holiday trip to NYC.

Aviv Tamar shiemannor Consider connecting to the automated experiment community. It's relatively small data, but rich in physical models that are well predictive. For example, arxiv.org/abs/2210.06526

Last finishing touches before the beginning of the new Computational Control course at ETH Zurich!

With Real-time Neural MPC you can efficiently integrate large, complex neural network architectures as dynamics models in an MPC-pipeline. Compared to prior implementations we can leverage neural networks with a 4000x larger parametric capacity in a 50Hz real-time framework.

He rocks

State-space models (SSMs) like Transformers are all the rage right now. Can these models implement classic algorithms from control? In joint work with Peter Bartlett, we show that how to implement a Kalman Filter and an LQG controller. 1/ arxiv.org/abs/2312.06937

The lab has 4 papers at ICRA this year IEEE ICRA: 1) TinyMPC: Model-Predictive Control on Microcontrollers 2) SLoMo: Legged Robot Motion Imitation from Casual Videos 3) VINSat: Visual-Inertial Navigation for Satellites 4) ReLU-QP: A GPU-Accelerated QP solver for MPC

Banquet with amazing people from TUM #ICRA2024

Eugene Vinitsky 🍒🦋 To tune each environment with grace, Use running mean, maintain the pace. Standard deviation, hold it tight, Normalizes returns just right. Clamp observations, rewards too, Within a range that's tried and true. Beware the bugs that physics send, A skewed value can offend. A

Nearly 80 distinguished speakers at #ICRA40! icra40.ieee.org/icra-2024/dist… Register by 29 August for early bird rates ⬇️ icra40.ieee.org/icra-2024/regi…

The original RL algorithms, inspired by natural learning, were online and incremental—they were streaming in the sense that they learned from each increment of experience as it happened, then discarded it, never to be processed again. The streaming algorithms were simple and

Unitree B2-W Talent Awakening! 🥳 One year after mass production kicked off, Unitree’s B2-W Industrial Wheel has been upgraded with more exciting capabilities. Please always use robots safely and friendly. #Unitree #Quadruped #Robotdog #Parkour #EmbodiedAI #IndustrialRobot

"Gr00t" vs "Pi0" vs "Pi0 Fast". I compared top open-source robotic models, and here's a detailed overview based on our own experience:

Stuck in local minima? We introduce tensor sampling + splines for high-entropy, smooth control trajectories, enabling high-entropy sampling-based MPC 🚀 Preprint: arxiv.org/abs/2505.01059 Website: sites.google.com/view/tensor-sa… Code: github.com/anindex/mtp #robotics #MPC #control