Griswald Brooks @simulcrashell

#PlotJuggler beta 2 released [1] There are a lot of bug fixes and improvements, but let's start with the highlight of this release: the Extension Marketplace! - Install/update plugins, without reinstalling the full app. - Contribute to the Marketplace with your own extension, using our SDK [2] - Isn't your file format supported yet. Contact me Also, the Windows installer is finally available!

China open-sourced a model that reconstructs any scene in 3D from a regular video, in real-time. one camera. no LiDAR. 10,000+ frames without falling apart. just walk around with your camera and watch the entire world get rebuilt in 3D at 20 fps. → runs at ~20 FPS on a single GPU → Stable over 10,000+ frames → Beats optimization-based methods on benchmarks → Works on drone footage, driving videos, indoor walkthroughs 100% open source.

We just hit a weird milestone: our model became more reliable than your average home WiFi. Just like everybody else, we thought cloud inference was the obvious choice. Yet 2 days into the ACT-2 eval, our mind completely changed. If our hero @ArpitKalla didn’t cook, this video wouldn’t exist 🧵

Introducing ACT-2 Preview The first robotics model to unify broad generalization with high reliability. A single fine-tuning example can teach Memo a new behavior that generalizes. Zero shot, real unseen homes, 99% success rate.

This is a really cool video of mobile manipulation

We found a general recipe for Solves: scale pretraining, then hill-climb with minimal in-house data. For the first time, one fine-tuning example can teach a new behavior that generalizes. Below: 4 folding strategies, learned from one example each and tested on held-out setups.

🎯 One CLI to evaluate every VLA benchmark Evaluating a robot policy used to mean wrangling six different repos, six different setups. Not anymore. LeRobot 0.6.0 ships 6 new simulation benchmarks - all runnable through the same lerobot-eval CLI, each with its own docs page, Docker image, and SmolVLA baseline tested in CI: 🧩 LIBERO-plus - ~10,000 perturbed variants across 7 axes, tells you exactly when a policy breaks 🤲 RoboTwin 2.0 - 50 bimanual tasks on SAPIEN + 100k+ ready-to-train trajectories on the Hub 🍳 RoboCasa365 - 365 kitchen tasks across 2,500 procedurally generated kitchens 🧠 RoboCerebra - long-horizon episodes chaining 3–6 sub-goals 🔁 RoboMME - a memory exam: count, track, imitate 📚 VLABench - knowledge & reasoning, from physics to brewing coffee end to end Add LIBERO, Meta-World, and NVIDIA IsaacLab-Arena, and that's 9 benchmark families under one roof - with parallel eval now up to 2x faster. See it in action below, and dive into the full release: huggingface.co/blog/lerobot-r…

Another awesome cheap robot

Antoine 🤖 @antoinemarcel

a week ago

Finally working :) The 1st home robot that's really easy to set up (< 30min for your 1st task), community-driven (if someone teaches it how to grab a beer and bring it to them, it’ll be able to do it too), and low-cost. Shipping 50 units soon. Link below if you want to join the

40 24 291 47K 114

Peter Corke won the highest honor in robotics. His entire robotics curriculum is free online. 📌 Has been for 10 years. 👇️ You visit his website and find 200+ free lessons, an open-source toolbox used by researchers worldwide, and a full textbook that runs in Google Colab. Zero install. He broke his knee before semester started in 2015. Recorded his lectures at home. Put them on YouTube. 70,000 views in one semester. He never stopped. → 200+ free robotics and robotic vision lessons → Full Python textbook... runs in Google Colab → Open-source Robotics Toolbox for research and teaching → Taken by millions of students globally From kinematics to deep learning. From arm robots to autonomous systems. Free. No login. [robotacademy.net.au] ——- Weekly robotics and AI insights. Subscribe free: 22astronauts.com

How do you know your robot actually succeeded? In v0.6.0, you ask a VLM 🤖 LeRobot ships a new unified reward models API with two zero-shot additions. Robometer is a general-purpose reward model trained on 1M+ trajectories across 21 robot embodiments: point it at any LeRobot dataset and it scores task progress and success straight from video and a language instruction, no task-specific training needed. TOPReward goes even further: no reward weights at all. It wraps an off-the-shelf VLM and reads the log-probability of the token "True" given the video and the instruction. Both come with labeling scripts that write per-frame progress curves into your dataset, ready for reward-aware behavior cloning. More on how they work: huggingface.co/blog/lerobot-r…

My company has rebranded from WTWH Media to Arrowfly, reflecting its focus on providing the best B2B content, events, and networks! You'll still find our #robotics coverage at The Robot Report and Automated Warehouse, and I'll see you at RoboBusiness! bit.ly/4wCnD4l

MolmoAct2 runs zero-shot on an SO-ARM101, no training required 🦾 @allen_ai open Action Reasoning Model is now in LeRobot with the full lifecycle covered: fine-tuning (full or LoRA), evaluation, and real-robot deployment. It pairs a Molmo2-ER vision-language backbone with a flow-matching action expert to turn images, language, and proprioceptive state into action chunks. The best part: ready-made checkpoints ship with calibration correction baked in, so you can point it at your SO-100/101 and just watch it work. Inference fits in ~12GB at bf16, and LoRA fine-tuning runs on a single 24GB GPU. Big thanks to the Ai2 team for building this in the open.

This is an actual, pretrained-from-scratch robot foundation model, something you basically never see. Most labs just start from a VLM or a pretrained video model. Awesome that its open source.

Yinghao Xu @YinghaoXu1

2 weeks ago

🚀 LingBot-VA 2.0 is here! After half a year of teamwork, we're thrilled to release LingBot-VA 2.0 — a native video-action foundation model for generalizable robot control. Unlike prior world-action models that retrofit generic video generators for robot control, LingBot-VA 2.0

20 67 458 82K 282

Meet VLA-JEPA, the first world model policy in LeRobot 🌍 Most VLAs just learn observation → action. VLA-JEPA also learns action-relevant dynamics: during training, a JEPA world model has to predict upcoming frames in latent space from the model's own actions. Then, at inference, the world model disappears entirely, leaving a standard, fast VLA: Qwen backbone, action head, nothing extra. In our demo it was fine-tuned on just 13 examples and ran in real time on an NVIDIA DGX Spark. On a consumer RTX 3080, it hits 10Hz using under 6GB of VRAM. World-model supervision during training, zero extra cost at inference. That's the whole trick.

This is #PlotJuggler Have you ever wanted to compare side by side two or more datasets simultaneously? Now you can! The new timeline editor allows you to align the time of different datasets. In the video, you can see two experiments executed days apart.

x.com/i/article/2073…

Real to sim is the true frontier for people obsessed with simulation

Yuke Zhu @yukez

3 weeks ago

Today, we're introducing SimFoundry, our real2sim2real framework at NVIDIA GEAR that automatically turns real-world scenes into simulation-ready worlds from a single image or video. Website: research.nvidia.com/labs/gear/simf… Paper: arxiv.org/abs/2606.28276 This work marks a major step for

16 74 491 48K 329

Today a guy told me "Compressed Depth image is very slow in ROS" and I answered "old by espresso". 3 hours later, I opened a Pr in the official image_transport_plugins repo. #depthz: 10x faster than PNG, much better compression.

They're getting better all the time even if right now it's a little like watching kids soccer

Shenzhen Channel @sz_mediagroup

4 weeks ago

Feel the robot “World Cup” in Shenzhen! From June 25 to 27, a robotics carnival in Shenzhen’s Longgang District brings thrilling robot football matches to the city. Robot football is gaining momentum across China, and Shenzhen is right at the center of the action. Come see it for

0 2 20 6K 3

@facontidavide Yeah, what now?