Eric Heiden (@eric_heiden) header
Eric Heiden (@eric_heiden) profile picture

Research in robotics and simulators @NVIDIA. Building Newton https://xitter.gq/t.co/q7OoFm5id6.

Joined March 2011
Photo by Eric Heiden (@eric_heiden)Photo by Eric Heiden (@eric_heiden)Photo by Eric Heiden (@eric_heiden)Photo by Eric Heiden (@eric_heiden)Photo by Eric Heiden (@eric_heiden)Photo by Eric Heiden (@eric_heiden)Photo by Eric Heiden (@eric_heiden)Photo by Eric Heiden (@eric_heiden)Photo by Eric Heiden (@eric_heiden)Photo by Eric Heiden (@eric_heiden)
The new solver Steel is now available in Newton!
Today we’re open-sourcing Steel, our internal physics solver built with @NVIDIA. It runs simulations up to 13× faster than PhysX while using ~3× less GPU memory. Steel unlocks scaleable simulation for manipulation, for everyone.
11
357
Eric Heiden retweeted
Today we’re open-sourcing Steel, our internal physics solver built with @NVIDIA. It runs simulations up to 13× faster than PhysX while using ~3× less GPU memory. Steel unlocks scaleable simulation for manipulation, for everyone.
57
142
70
1,480
132,017
We’re excited to announce the official release of NVIDIA Isaac Lab 3.0. 🎉 Isaac Lab 3.0 brings a new modular and lightweight architecture along with major new capabilities across the robot learning workflow: multi-physics and multi-rendering support, training at scale, and streamlined deployment on real hardware. With Newton integration, the new OVRTX renderer, RL post-training, and LEAPP deployment, there’s a lot to explore. Here’s a look at what’s new 🧵 (1/7)
46
187
49
1,417
102,165
Eric Heiden retweeted
Excited to release the RoboSimStudio toolkit today to help agents create detailed dynamics Real2Sim for Agentic Robotics.
We gave Claude and GPT a phone video of a robot wiping. In an hour, each built a full-dynamics digital twin. Introducing RoboSimStudio 🤖: an open and interactive toolkit for Agentic Robotics, enabling Fable, Astra and Argon AI agents to model sponges 🧽, cables 🧵 and cloth 👔.
6
17
3
164
112,767
Newton 1.0 is now generally available. 🙌 Take robot learning to the next level with: 🤖 Stable Articulated & Complex Mechanism Simulation – accurate, reliable machine modeling. 🖐️ High-Fidelity Hydroelastic Contact Modeling – realistic soft contact and touch-based interactions. 🧵 Deformable Body Simulation – simulate cables, cloth, rubber, and other elastic materials with VBD. ⚡ Accelerated Robot Learning at Scale – seamless integration with open simulation and learning frameworks, NVIDIA Isaac Sim and Isaac Lab for scalable workflows. Learn how to integrate this open-source physics engine into your workflow: nvda.ws/3NGTzUo #NVIDIAGTC
25
148
14
1,062
83,058
Eric Heiden retweeted
For all the MuJoCo fans: MimicKit now supports Newton / MuJoCo Warp! Fast, lightweight MuJoCo physics with Warp acceleration. github.com/xbpeng/MimicKit/ Coming up: God's physics engine
11
94
4
819
54,102
🚀 Newton Beta 2 and roadmap to 1.0 is out. SDF-based collisions & hydroelastic contacts for manipulation, cable-type deformables in VBD solver, stronger MuJoCo Warp integration and a new L-BFGS optimizer for IK solver are among major features of Beta. github.com/newton-physics/ne…
4
4
626
We are honored that our NVIDIA Warp has received two @HPCwire Readers’ Choice Awards this year. 🏆 Best HPC Programming Tool or Technology 🏆 Top 5 New Products or Technologies to Watch NVIDIA Warp, an #opensource CUDA-X library, accelerates computational physics by up to 245x. 📗 Learn how Warp can transform your workflow: developer.nvidia.com/warp-py… 📝 Dive deeper into the latest release: github.com/NVIDIA/warp/relea… #HPCwireRCA25 #SC25
2
14
49
4,718
A robot that learns in its dreams - and then those dreams come true. 🤯 It studies physics until it no longer needs equations, just intuition. They call it NeRD: Neural Robot Dynamics. The student that beat the teacher. What a time to be alive! Full video: youtu.be/M8s_cS-aH5w
1
16
2
71
22,620
Good summary of our paper on neural physics - fine tuning learned simulators on real world data is indeed a key motivation. Another nice property is that now your physics engine can run on Tensor Cores and can benefit from all the hardware NVIDIA develops to make LLMs go fast.
Why would anyone need «neural physics»? Well, at least one problem it solves is the sim-to-real gap, and that’s what this paper is about NeRD is a neural simulation module that can predict future states of articulated rigid bodies, completely replacing time integration and solvers Initially, it’s pre-trained on classic simulation data, which is cool, but doesn’t add much value by itself The important part: it can be fine-tuned on real data to handle wear and tear, contact reality, and environmental changes, significantly improving dynamics accuracy compared to the analytical simulator The architecture is surprisingly simple, it’s basically a GPT-2 with a tiny history window and a robot-centric tokenization That design decision makes it fast too, faster than the analytical solution arxiv.org/pdf/2508.15755
3
25
3,180
Let's hear it for the community. 🎉 The Newton project, an open-source physics engine for robotics simulation, is now part of the @linuxfoundation. This is a huge step toward building a truly community-driven future for robotics research and development.
Newton is now part of the Linux Foundation! 🤖Co-developed by Disney Research, Walt Disney Imagineering, @GoogleDeepMind & @nvidiarobotics, the open source GPU-accelerated physics engine helps robots learn faster, scale safely, and transfer skills to the real world. Learn more about Newton: hubs.la/Q03Lpt2S0
5
14
1
42
6,406
The laws of physics apply everywhere. ⚛️ Co-developed by NVIDIA, @GoogleDeepMind and Disney Research, Walt Disney Imagineering, the new Newton Beta, managed by the @linuxfoundation, now also runs with Isaac Lab and with MuJoCo Warp, which delivers warp speed for robot learning. Learn how to train a quadruped and simulate multiphysics in our tech blog. 🔗 nvda.ws/4gP2woL #CoRL2025
23
122
13
783
83,246
🤖🤖🤖 Following RoboVerse, we introduce another work focused on Robotic Tactile Simulation - Taccel Simulator. Taccel is a high-performance simulation platform for vision-based tactile sensors and robots. 🚀🚀🚀 Boosted by Nvidia Warp, we optimize Taccel with highly parallelized simulations and support 900fps simulation with 4k+ parallel training envs. 🤝🤝🤝 Taccel is designed with user-friendly APIs and is easy to use. We open-sourced all the code and documentation. Feel free to try! Project: taccel-simulator.github.io/ Preprint: arxiv.org/abs/2504.12908 Code: github.com/Taccel-Simulator/…
3
33
7
137
10,738
📣MuJoCo announcement 📣 Thrilled to share that @GoogleDeepMind has unveiled MuJoCo-Warp at @nvidia's #GTC25! 🚀 We've expanded our open-source MuJoCo simulator with MuJoCo-Warp, leveraging NVIDIA’s Warp framework for incredible acceleration. This marks a significant step in making high-performance simulation more accessible.
18
87
4
608
82,282
Big news from Jensen at GTC: NVIDIA Warp is now open source under Apache 2.0! We’re also building Newton, a new physics engine for robotics with DeepMind and Disney—based on Warp and fully open-source.
Btw... Warp just transitioned to an Apache 2.0 license! 👀 developer.nvidia.com/warp-py…
4
41
268
26,315
Register for the @GoogleDeepMind session at #GTC25 to learn how Google integrates core NVIDIA technologies like Warp to improve simulation performance for #humanoid robotics development. Register Now ➡️ nvda.ws/3Xya1YM
2
21
8
128
30,594
NVIDIA Warp 1.6.0 is out! Including native Blackwell support, up to 2x faster geometric queries, and tile-based Cholesky decompositions. Full CHANGELOG here: github.com/NVIDIA/warp/blob/…
1
35
1
178
11,272