🚀 If you’re looking for a batchable, differentiable, and fully GPU-accelerated MPC for your learning and control applications, check out our latest work! We just published: 📄 "Primal-Dual iLQR for GPU-Accelerated Learning and Control in Legged Robots" in IEEE Robotics and Automation Letters (2025). Our solver is written entirely in JAX, supports automatic differentiation, and runs in massively parallel environments—a useful feature when coupled with IsaacLab/GPU-based simulators for RL and Imitation Learning! 🔥 Key benefits: ✅ Long prediction horizons: WB-MPC up to 300 nodes in <25 ms ✅ Centralized control: WB-MPC for up to 4 robots in <25 ms ✅ Massive parallelization: 35× real-time with 4096 simulated environments 📄 Paper (IEEE RA-L 2025): https://lnkd.in/dA6Qsahz 💻 Open-source code: https://lnkd.in/ddAmKFrX Authors: Lorenzo Amatucci*, João Sousa-Pinto*, Giulio Turrisi, Dominique Orban, Victor Barasuol, and Claudio Semini (*Equal contribution) #Robotics #MPC #GPUComputing #JAX #DifferentiableControl #LeggedRobots #Optimization #ControlSystems
Dynamic Legged Systems Lab
Research Services
Genoa, Liguria 2,853 followers
The DLS lab is performing research + development for the design + control of high-performance, versatile legged robots.
About us
The Dynamic Legged Systems focuses on research that concerns the design and control aspects of agile legged robots. We are interested in the development of legged robotic systems, chiefly quadrupeds, and we are investigating ways that can increase flexibility and performance of legged designs. Apart from developing agile quadruped robots we emphasize the need for accurate and robust control, ranging from the level of individual joints up to the overall behavior of the legged robot. Apart from accurate hydraulic force/torque control at the joint level we are investigating ways of creating and using a variety of different locomotion gaits, that are robust to external disturbances and changing environmental conditions. A significant part of our groups' effort is devoted to perception tailored to dynamic quadrupedal robots. Our flagship platforms are the HyQ robots (HyQ, HyQ2Max, HyQReal). Since its construction in 2010, HyQ demonstrated a wide repertoire of motions ranging from stair climbing, chimney climbing, omnidirectional trotting over rough terrain, trotting with step reflexes, to flying trotting and jumping. Our latest generation quadruped HyQReal has been released in May 2019 demonstrating its strength by pulling a small passenger airplane.
- Website
-
https://dls.iit.it
External link for Dynamic Legged Systems Lab
- Industry
- Research Services
- Company size
- 11-50 employees
- Headquarters
- Genoa, Liguria
- Type
- Educational
- Founded
- 2007
- Specialties
- Robotics, Hydraulics, Locomotion, Planning, Perception, Software, Machine Learning, Mechatronics, Robot Design, Agricultural Robotics, Robots in Space, and Construction Robots
Locations
-
Primary
Get directions
Via San Quirico, 19D
Genoa, Liguria 16163, IT
Employees at Dynamic Legged Systems Lab
-
Giulio Turrisi
Postdoctoral researcher @ Istituto Italiano di Tecnologia
-
Saeed Abdollahi
Robotics Software Engineer | Specializing in Perception and Sensor Fusion
-
Lorenzo Amatucci
PhD candidate at Dynamic legged system, IIT
-
Marco Marchitto
Robotics Software Engineer at Istituto Italiano di Tecnologia - Dynamic Legged System Lab
Updates
-
What if your robot had to nagivate on "No-Mercy" terrains, like on collasping surfaces where a single misstep could represent a catastrophic failure? In our recently accepted IEEE RA-L paper, we developed a novel locomotion framework for safe navigation over unstable surfaces by integrating terrain probing, load-bearing analysis, motion planning, and control strategies relying purely on joint measurements to assess terrain stability. Title: Load-bearing Assessment for Safe Locomotion of Quadruped Robots on Collapsing Terrain Authors: Vivian Suzano Medeiros, Giovanni Battista Dessy, Thiago Boaventura, MARCELO BECKER, Claudio Semini, and Victor Barasuol Affiliations: IIT Dynamic Legged Systems Lab and USP EESC. Link to the paper: https://lnkd.in/dMWsHWbK Video: https://lnkd.in/dKJd35Fh #leggedrobot #locomotion #Robotics #QuadrupedRobots #MotionPlanning #MPC
Load-bearing Assessment for Safe Locomotion of Quadruped Robots on Collapsing Terrain (RA-L, 2025)
https://www.youtube.com/
-
Dynamic Legged Systems Lab reposted this
Call for extended abstracts for Space Robotics workshop at IEEE iSpaRo 2025, Sendai, Japan (1-4 December 2025). Workshop: AI4OPA: AI Techniques and Frameworks for Orbital and Planetary Autonomy Workshop date: 1/December/2025 Submission deadline for extended abstracts: 10/NOV/2025 https://lnkd.in/dRGKh9Uv The #AI4OPA workshop will explore how Artificial Intelligence (AI) can serve as powerful tools to enhance the autonomy of both orbital and planetary space missions. A dedicated subtopic of the workshop is the open-sourcing of space-related datasets and benchmarks that are essential for the training and evaluation of Machine Learning models. Organizers Matteo El Hariry, University of Luxembourg Lisa Denzer 🚀👩🏻🚀🛰️, European Space Agency Simon Bøgh, Aalborg University Claudio Semini, IIT Genova Teresa Vidal Calleja, UT Sydney Amir Rahmani, NASA JPL #SpaceRobotics #iSpaRo #AI #AI4OPA #Space #robotics #conference #Learning #orbital #planetary #autonomy
-
🚀 The DLS team will present three RA-L papers at IROS 2025 in Hangzhou, China! If you’re attending, come see our talks and chat with us about our latest work on legged locomotion, perception, and state estimation 👇 📅 Wednesday, Oct 22 🕐 13:35–13:40 — Sensor Fusion 6 session, room 102C 🔹 “Multi-Sensor Fusion for Quadruped Robot State Estimation Using Invariant Filtering and Smoothing” Authors: Ylenia Nisticò*, Hajun Kim*, João Carlos Virgolino Soares, Geoff Fink, Hae-Won Park, Claudio Semini (*Equal contribution) 👉 Paper link: https://lnkd.in/d2EYSCBQ 🕔 17:10–17:15 — Legged Robots 4 session, room 103A 🔹 “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) 👉 Paper link: https://lnkd.in/dG-TsAPr 📅 Thursday, Oct 23 🕐 13:25–13:30 — Legged Robots 6 session, room 103A 🔹 “MUSE: A Real-Time Multi-Sensor State Estimator for Quadruped Robots” Authors: Ylenia Nisticò, João Carlos Virgolino Soares, Lorenzo Amatucci, Geoff Fink, Claudio Semini 👉 Paper link: https://lnkd.in/d7se6kdA We’re looking forward to connecting with the robotics community in Hangzhou — see you at #IROS2025!
-
We're #hiring a new Test Engineer for Space Robotics in Genoa, Liguria. Apply today or share this post with your network.
-
Claudio’s visit at the Space Robotics Laboratory in Sendai, Japan!
Last week I had the pleasure to visit the SRL - Space Robotics Laboratory at University of Tohoku and give a seminar. I enjoyed the impressive lab tour, the cool robots (I can’t post pics of some of them yet), and the interaction with the team members! I would like to thank Shreya Santra, Ph.D and Kentaro Uno for hosting me and all the lab members for their time during the lab tour! Yoshida sensei, I hope our agendas will match better next time to meet again. #robotics #space #spaceexploration #MoonShot #modular #HubRobo #SRL #TohokuUniversity #Japan
-
-
Dynamic Legged Systems Lab reposted this
Il Festival dello Spazio non si racconta soltanto: si vive. A Villa Borzino potrai osservare un robot quadrupede affrontare un “terreno lunare”, scoprire una serra adattiva progettata per ambienti extraterrestri, e vivere in prima persona il brivido del Launch Pad e delle maschere subacquee usate dagli astronauti NASA - National Aeronautics and Space Administration. 🔬 A cura di: 🤖 Claudio Semini – Istituto Italiano di Tecnologia 🌱 Bianca Simonassi – Space V 🚀 Pietro Bazzani – Space V La ricerca prende forma e diventa dimostrazione. #festavaldellospazio #festivaldellospaziobusalla #festivaldellospazio2025
-
Open PhD positions at the DLS Lab! The Dynamic Legged Systems lab has two open PhD positions: • Learning-based Manipulation Skills for Agricultural Robotics • Quadruped Robot Control for Environmental Sustainability using Machine Learning All positions are located in Genova at the Istituto Italiano di Tecnologia (IIT) and are fully funded. For more details about the application process and the positions visit: https://lnkd.in/dYF2S_cm The application deadline (compulsory submission through the UniGE website https://lnkd.in/gqYrRRtq) is 9th July 2025, 12.00 noon (CEST). Istituto Italiano di Tecnologia #PhD #Robotics #Quadrupeds #iitDLSlab #IIT #DLS #IRIM #UniGE #AgriculturalRobotics #EnvironmentalSustainability
-
-
VERO @ Green and Blue Festival 2025!! Today and tomorrow, Saturday, our team is in Milan at the Museo Nazionale Scienza e Tecnologia Leonardo da Vinci, showcasing VERO, the quadruped robot powered by AI, to help remove cigarette butts littered in the environment. (https://lnkd.in/dzMriEcq) Come meet us, and check out our system in action!🤖 #VERO #Robotics #Innovation #Sustainability #LeggedRobotics #Quadrupeds #Fieldrobotics #RaiseLiguria
-
-
If you are attending the IEEE International Conference on Robotics and Automation (ICRA), don't miss the presentation of our paper "SANDRO: a Robust Solver with a Splitting Strategy for Point Cloud Registration", given by João Carlos Virgolino Soares! Session: Point Cloud Registration - Thursday 22, 8:50-8:55 AM, GMT-4
We are thrilled to announce that our latest research, "SANDRO: a Robust Solver with a Splitting Strategy for Point Cloud Registration", has been accepted to the 2025 IEEE International Conference on Robotics and Automation (ICRA). In this work, we introduce SANDRO, an innovative algorithm that enhances point cloud registration by combining an Iteratively Reweighted Least Squares (IRLS) framework with a robust loss function and a novel splitting strategy. This approach effectively addresses high outlier rates and skewed outlier distributions, leading to significant improvements in registration success rates. Notably, SANDRO demonstrates a 20% improvement over current state-of-the-art methods on the Redwood real dataset and a 60% improvement on synthetic data. We invite you to explore our findings in detail in the full paper. Authors: Michael Adlerstein, João Carlos Virgolino Soares, Angelo Bratta, Claudio Semini Read the full paper here: https://lnkd.in/dASM4CHE Our code will be available at https://lnkd.in/dk4HwVYs during the ICRA 2025 conference!
-