Giulio Turrisi

Ping everyone

MuJoCo Playground An open-source framework for GPU-accelerated robot learning and sim-to-real transfer

Introducing playground.mujoco.org
Combining MuJoCo’s rich and thriving ecosystem, massively parallel GPU-accelerated simulation, and real-world results across a diverse range of robot platforms: quadrupeds, humanoids, dexterous hands, and arms.
Get started today: pip install playground

Title: Non-Gaited Legged Locomotion with Monte-Carlo Tree Search and Supervised Learning
Authors: Ilyass Taouil*, Lorenzo Amatucci*, Majid Khadiv, Angela Dai, Victor Barasuol, Giulio Turrisi, Claudio Semini
*Equal Contribution
Link to the paper: arxiv.org/abs/2408.07508

Non-Gaited Legged Locomotion with Monte-Carlo Tree Search and Supervised Learning (RA-L 25) YouTube video by Dynamic Legged Systems lab

The combinatorial nature of contact planning in legged robots hinders the success of optimal control in navigating complex scenarios. In our recent RA-L paper, we demonstrate how sampling-based methods and supervised learning techniques can be coupled to search for a solution in real-time.

IEEE RA-L enters its double-blind era

LOT Spring School Topics

🐱 Learning Over Time (LOT) Spring School
📅 24-27 March 2025 | 📍 Siena, Italy

💡 Are you considering models that continuously adapt over time instead of learning "offline" from pre-designed-huge collections of data? 🚀

🔗 sites.google.com/unisi.it/lot...

#ContinualLearning #CL #LifelongLearning

E61: Neurips 2024 RL meetup Hot takes: "What sucks about RL?"
What do RL researchers complain about after hours at the bar?  In this "Hot takes" episode, we find out!  
Recorded at The Pearl in downtown Vancouver, during the RL meetup after a day of Neurips 2024.

Yesterday the hyped Genesis simulator released. But it's up to 10x slower than existing GPU sims, not 10-80x faster or 430,000x faster than realtime since they benchmark mostly static environments

blog post with corrected open source benchmarks & details: stoneztao.substack.com/p/the-new-hy...

Thanks for the insight! Looking forward to the report!

GitHub - Genesis-Embodied-AI/Genesis: A generative world for general-purpose robotics & embodied AI learning. A generative world for general-purpose robotics & embodied AI learning. - Genesis-Embodied-AI/Genesis

The next isaac sim/lab?

github.com/Genesis-Embo...

I'm looking forward to a better cpu in the lineup! Maybe the novelty that apple brought in ARM can be go there as well.

It is a nice idea! Wondering why they chose to go for 4+1 pages instead of 6+. This can make people considering it not a "full" paper

Reinforcement Learning: An Overview This manuscript gives a big-picture, up-to-date overview of the field of (deep) reinforcement learning and sequential decision making, covering value-based RL, policy-gradient methods, model-based met...

An updated intro to reinforcement learning by Kevin Murphy: arxiv.org/abs/2412.05265! Like their books, it covers a lot and is quite up to date with modern approaches. It also is pretty unique in coverage, I don't think a lot of this is synthesized anywhere else yet

Authors: Giulio Turrisi, Lucas Schulze, Vivian S. Medeiros, Claudio Semini, Victor Barasuol
Paper: arxiv.org/pdf/2403.19862

(Other Twitter stuff)

Human-robot and robot-robot collaborative carrying with mobile robots typically requires a manipulator arm. At iros2024, we introduce PACC, a novel passive-arm design that, coupled with a decentralized MPC, allows an efficient execution of such tasks with quadruped robots.

(Posting old Twitter stuff for boosting connections/the algorithm)

Open-source (and in python!) MPC for quadruped robots. Gradient-based (via acados) or sampling-based in #JAX. Plus, multiple robots, multiple terrains, and multiple gaits!

Link: github.com/iit-DLSLab/Q...

Very nice extension!

Not yet, but we are working on it! If you discretize too much, the consensum mechanism will be slower to converge (and the will be more errors in the meantime). So yes, there is a sweet spot, but one can play a lot with the optimal subdivision here

Resercher in legged locomotion from italy here!

Title: Accelerating Model Predictive Control for Legged Robots through Distributed Optimization
Authors: Lorenzo Amatucci , Giulio Turrisi , Angelo Bratta , Victor Barasuol , Claudio Semini
Link to the paper: arxiv.org/pdf/2403.11742
Code: github.com/iit-DLSLab/D...

2/2

Model Predictive Control (MPC) for legged robots is limited by model complexity. In our recent IROS24 paper, we show how partitioning the robot into smaller subsystems, each controlled in parallel by a faster dedicated MPC, can achieve complex whole-body motions more efficiently.

1/2

precise motion that requires a lot of planning are still out of reach (e.g. cat agility!)