Mac Schwager
@MacSchwager
Followers
667
Following
202
Media
0
Statuses
93
I build robots that safely and intelligently interact with each other and with humans. I am a professor at Stanford, where I lead the Multi-robot Systems Lab.
Joined December 2020
Do world models for robots really need to predict RGB videos? Probably not! Our ParticleFormer shows that 3D structure, semantics, and fine grained action conditioning all seem to be more important than RGB appearance. Congrats to Suning and an amazing team on this work!.
🚀 Excited to share our #CoRL2025 paper! See you in Korea 🇰🇷!🎉. We present ParticleFormer, a Transformer-based 3D world model that learns from point cloud perception and captures complex dynamics across multiple objects and material types !. 🌐 Project website:
1
14
99
Can RL train visual motor policies for drone navigation? Yes! With lots of help from high fidelity 3DGS visuals and end2end differentiability from images to control inputs. Congrats to @SrRjHS1CSi1EGL3 and an amazing @StanfordMSL team on this work, just accepted to IROS.
[1/2] Excited to announce GRaD-Nav! We propose a new framework that integrates 3DGS and Differentiable RL to train vision-based drone navigation policies. Our method achieves efficient end2end training, zero-shot sim2real transfer, and strong in-task adaptability.
0
1
10
Generate videos in just a few seconds. Try Grok Imagine, free for a limited time.
394
661
3K
RT @tennyyin: 🔎Can robots search for objects like humans?.Humans explore unseen environments intelligently—using prior knowledge to activel….
0
49
0
For robots to learn to collaborate with other robots and humans, they need to predict others' latent states. They need a latent theory of mind. 👇.
[1/5] Humans collaborate with each other by simulating the state of mind of their teammates, a concept called Theory of Mind (ToM). We propose LatentToM, a method to endow robots with a theory of mind in latent space for cooperative manipulation.
2
5
24
Robot diffusion policies are so powerful because of their awesome action generalization capabilities. right??? Nope, quite the opposite.👇.
[1/5] What happens when you prompt a robot diffusion policy with an image of a cat? .Website: Paper:
1
1
25
RT @das_princeton: (1/13) How should we rigorously compare robot policies? Comparison is central to robotics research, but is inherently ex….
0
7
0
Code, training data, and our GSplat sim environment (we call Flying in Gaussian Splats, FiGS) are all open source. You can train your own visual-nav policy for your own environment, even without a drone. Just upload your images, train a GSplat, and you're good to go.
0
0
0
Sous Vide (👩🍳): cooking food to perfection in a vacuum sealed bag. SOUS VIDE (🤖): cooking visual-motor robot policies to perfection in a Gaussian Splatting scene model. Both yield delicious results. Very proud of JunEn Low and the @StanfordMSL team for this work.
[1/3] We're excited to announce SOUS VIDE, a method to train visuomotor navigation policies for autonomous drones without data from real-world flights, using only Gaussian Splat reconstructions of the scene. 📄 Paper: .🌐 Project:
2
1
12
RT @janusch_patas: SIREN: Semantic, Initialization-Free Registration of Multi-Robot Gaussian Splatting Maps. Contributions:.• We introduce….
0
3
0
Stanford Aero-Astro is hiring! We have an open-rank faculty search this year. Find the full job posting here: .
0
0
5
Can dexterous multi-fingered policies transfer from sim to real? Yes!. using grasp evaluators trained on a big enough dataset. 👇.
There have been many recent big grasping datasets, but few demos of real-world grasping using generative models. How do we achieve this?. Introducing: Get a Grip (#corl2024)!. We show that instead of generative models, discriminative models can attain sim2real transfer!. 👀🧵👇
0
1
11
Want a modular, interpretable alternative to VLAs? Check out Splat-MOVER, which uses language embedded, editable GSplats for open vocabulary multi-stage manipulation. Congrats to Ola Shorinwa, John Tucker, and the rest of an amazing team.👇.
Excited to announce Splat-MOVER for multi-stage, open-vocabulary manipulation, with:.- Semantic and affordance scene understanding.- Scene-editing.- Robotic grasp generation!. Find out more at and join us Friday, 11/08/24, at the #CoRL2024 poster session.
0
0
16
RT @tylerlum23: It was a pleasure working with🤝with.@albert_h_li, @pdculbert, @krishpopdesu, Aaron Ames, @MacSchwager, and @leto__jean!. 💻W….
0
2
0
RT @StanfordMSL: How much should you trust your robot? Check out our latest paper, where we place sample-efficient confidence bounds on the….
0
2
0
RT @janusch_patas: Touch-GS: Visual-Tactile Supervised 3D Gaussian Splatting.Project: Paper: .
0
12
0
RT @zhenjun_zhao: Splat-Nav: Safe Real-Time Robot Navigation in Gaussian Splatting Maps. Timothy Chen, Ola Shorinwa, Weijia Zeng, Joseph Br….
0
12
0
I'm very excited about this project surveying recent literature on foundation models in robotics. Congratulations to @royafiroozi on her leadership and vision in bringing the efforts of the awesome co-authors together into a comprehensive survey.
Check out our survey paper entitled "Foundation Models in Robotics: Applications, Challenges, and the Future," which spans 233 recent papers on this topic. ArXiv link: 1/5.
0
1
15
We use Conformal Prediction (CP) to obtain rigorous statistical performance guarantees for diffusion policies. We also include a CP-inspired quantile loss during training to further improve robustness of the policy. We call it PlanCP. Check it out at NeurIPS.
PlanCP quantifies the uncertainty of diffusion dynamics models using Conformal Prediction (CP). 📜 Paper: 🕗 Great Hall & Hall B1+B2 (Level 1), #438, at the New Orleans Ernest N. Morial Convention Center on Wed, Dec 13th, 5-7 p.m. CST. #NeurIPS2023
0
1
24
Stanford Aero-Astro has an open search for tenure-track junior level faculty. Please apply, and pass this on to friends and colleagues who may be looking. See Marco's tweet for links and details. 👇.
The deadline to apply for a faculty position @Stanford.Aero/Astro is next week, December 4, 2023. Hurry up if interested! More info here:
0
0
1
RT @drmapavone: The @Stanford Aero/Astro Department is hiring! The Department invites applications for a tenure track faculty position at t….
0
4
0