Robotics PhD student @ RSL | JPL | ETH | IIT Kgp
Zurich, Switzerland
Joined May 2019
- Tweets83
- Following406
- Followers447
- Likes2K
Pinned Tweet
Vision foundation models like DINO and SigLip now power much of robot learning.
But they are built for RGB. What about depth?
Introducing DeFM: a foundation model for depth images, built for robotics.
🌐 de-fm.github.io
Weights, code, and demo are out today 🧵
Manthan Patel ✈️ CoRL retweeted
Did this because I think AME-2 deserves a complete verification on humanoids. It might be a bit late tho.
Some sim perf: collision avoidance on sparse stepping stones, foot-only climbing of 70 cm obstacles, 1.2 m gap traversal (can be larger I guess)...
github update soon.
Manthan Patel ✈️ CoRL retweeted
🎯Demo is on the shelf!
wuyan01.github.io/ADAPT-proj…
Try out the language-conditioned end-to-end humanoid control in your browser!
And our paper:
📄 ADAPT: Agile Diffusion Action Priors for Robust and Steerable Online Text-Driven Humanoid Control
arxiv.org/abs/2609.00677
Dominant text-to-motion pipelines generate kinematic motions for a separate tracker.
Instead, in our @corl_conf work ADAPT, we solve humanoid language control end-to-end closed-loop, where the robot must continuously respond to changing commands.
🌐Page: wuyan01.github.io/ADAPT-proj…
Papers and project pages:
🧗 EgoHTR: egohtr.github.io
🧠 DeFM: de-fm.github.io
🗺️ WayTrace: way-trace.github.io
📏 VLD: leggedrobotics.github.io/rl_…
🌲 WildOS: leggedrobotics.github.io/wil…
📡 RangeNav: rangenav.github.io
Rooms and times are in the image ☝️
Manthan Patel ✈️ CoRL retweeted
My robot teleportation system was featured on Swiss National Television! SBB, the Swiss national railway company, is interested in teleoperated robots for railway tunnel maintenance.
Full article here: srf.ch/news/schweiz/projekt-…
Getting a step closer to long-range city-scale Navigation powered by Foundation Models 🚀
A sneak peek into what we have been building. Stay tuned for the release
p.s. Can you guess this popular location?
Full thread and paper now available
xitter.cf/patelm99/status/211798…
Manthan Patel ✈️ CoRL retweeted
Our new work WayTrace: an off-the-shelf VLM, an onboard trace model, and an RL policy take a robot from web-map directions to 2.35 km of real-world missions in Zurich.
Details in the thread 👇
We gave a robot the walking directions you'd get from a web map.
Then we let it find its own way through Zurich: crosswalks, stairs, crowded streets, construction sites, 2.35 km in total.
Introducing WayTrace 🧵👇
10/11
🌐 Project page: way-trace.github.io
📄 Paper: arxiv.org/abs/2611.05129
▶️ Full video: youtube.com/watch?v=FrujvU8m…
💻 Code, models, dataset and benchmark: coming soon
Work from the Robotic Systems Lab (@leggedrobotics ) at ETH Zurich (@ETH_en), with UC Berkeley (@UCBerkeley) and NVIDIA (@nvidia)
With @JonasFrey96, Pascal Roth, Marco Hutter, and Cesar Cadena
Manthan Patel ✈️ CoRL retweeted
This is one of the most promising depth encoders I’ve seen, demonstrating a remarkable ability to even enable segmentation directly from depth. Excited to see the many future works that build upon this encoder!
#Robotics #foundation #models #learning #vision #research
Vision foundation models like DINO and SigLip now power much of robot learning.
But they are built for RGB. What about depth?
Introducing DeFM: a foundation model for depth images, built for robotics.
🌐 de-fm.github.io
Weights, code, and demo are out today 🧵