Build on @nvidia IsaacSim with Gemini API through @antigravity. Middleware uses ROS2 to plan and execute pick-and-place action to the FR3 robot. Keywords: #Sim2Real, #DigitalTwin, #Simulation

Build on @nvidia IsaacSim with Gemini API through @antigravity . Middleware uses ROS2 to plan and execute pick-and-place action to the FR3 robot. #Robotics #Gemini #Antigravity #AI #ROS2

Bước 2GPU làm phép màu: -Trajectory được replay trên GPU Linux servers. -IsaacSim backend biến nó thành môi trường photorealistic. -Domain-randomized = robot học trong hàng nghìn hoàn cảnh khác nhau từ 1 trajectory duy nhất của bạn.

【自作 #ヒューマノイド TERABO-3 に向けて🤖】 二足歩行のための脚の設計🦵 CAD上では滑らかに動くものができてきました！ 中身はまだまだ進化します！🧠 #ロボット #ヒューマノイド #isaacsim #humanoid #robotics #terakatarobotics

Finally connected #IsaacSim running natively on #Windows with #ROS2 in #WSL2 (Ubuntu 24.04) on my home desktop PC! Bridging them locally is a total game-changer. Automated the entire setup & troubleshooting using a multi-LLM stack (#Gemini 3.5 Flash, 3.1 Pro, & #Opus 4.6 via Google AI Pro) alongside the Antigravity 2.0 CLI & IDE. Incredible performance! 🚀 Stay tuned for what I build on top of this next. 👇 @NVIDIARobotics @GeminiApp @antigravity

DM me if you are expert at MuJoCo/IsaacSim and want to be a simulation engineer.

Controlled a Maya 3D rig with a real robot arm — live. Moved the SO-101 leader arm, joint angles streamed over WebSocket through ROS 2 running in Docker #Robotics #RobotLearning #EmbodiedAI #Simulation #Omniverse #IsaacSim #IsaacLab #RTX #AI

You are sitting at home. A robot across the world just learned something new. Because of you. @axisrobotics makes this possible. No hardware. No lab. Just a browser. You control a simulated robot arm. Your session is recorded as verified training data. IsaacSim turns it into thousands of photorealistic samples. Real robots train on what you just did. Every contribution is registered on-chain. Your data has an ID. A value. An owner. Robot intelligence was never supposed to belong to one company. It belongs to the people who built it. @axisrobotics @base #PhysicalAI #Robotics #Web3 #DePIN #RoboticsAI

Let's build agentic-robotic-manipulator agent using @antigravity and @GeminiApp. Simulator is running on @NVIDIARobotics' #IsaacSim. They are working well on windows 10 and wsl2 with #ROS2

Agentic AI under #IsaacSim of @NVIDIARobotics . Seamless digital-twin vs Visualized digital-twin. Which one is your prefer?

camera setup #Robotics #RobotLearning #EmbodiedAI #Simulation #Omniverse #IsaacSim #IsaacLab #RTX #AI

SO-ARM101 has arrived 🤖 Got the arm up and running and completed my first teleoperation tests today. More updates coming soon. #Robotics #RobotLearning #EmbodiedAI #Simulation #Omniverse #IsaacSim #IsaacLab #RTX #AI

Hit 1 million RL steps/sec today! This policy trained for 1 minute on a single GPU using a from-scratch physics simulator built with custom Triton kernels. The sim runs at about 5 million physics steps per second. With decimation = 4, this produces ~1.3 million env steps/second. The PPO training loop brings it down to ~1 million steps/second. Modern GPU-accelerated physics engines like IsaacSim and MJWarp are amazing, but they are, by design, general purpose. For any specific problem, they can leave a lot of performance on the table. And once you start coloring outside the lines, throughput can fall apart fast. The video below is sim2sim verification: trained in our sim and evaluated against IsaacLab's built-in Go2 environment. Clean transfer!

🔹IsaacSim Augmentation: Turning single trajectories into thousands of photorealistic, domain-randomized training samples for VLA models.

X Post Idea (BitRobot × Axis Robotics) Physical AI doesn’t scale without high-quality data. With SN/04, @AxisRobotics joins the BitRobot Network to expand teleop-in-sim data collection, turning simple actions like grasping, placing, and rotating objects into valuable training data for next-generation robotics models. 🔹 1,200 robotic manipulation tasks 🔹 IsaacSim-powered data augmentation 🔹 Vision-Language-Action (VLA) model training 🔹 Double rewards for contributors through BitRobot & Axis Every successful demonstration helps robots better understand and interact with the physical world. The future of robotics isn't just about better models—it's about better data pipelines. 🤖 @BitRobotNetwork @axisrobotics #BitRobot #AxisRobotics #PhysicalAI #Robotics #AI #MachineLearning #Teleoperation #VLA #DePIN #ArtificialIntelligence

今日のIsaacLabのお勉強はここまで。 自作の２関節ロボの強化学習開始まではたどり着けた（スクリプトの設定とかむずいぞ！）。　明日は報酬設定を見直して、ちゃんと立つようになりますように。 #IsaacLab #IsaacSim

Who is gonna catch him first ? 🤣 #IsaacSim #MCP #AIAgents #Robotics #NVIDIA

Okay My team and I are building local industrial robot arm with emphasis on dexterity in Africa and delivery drones. Thing is we have Gazebo, IsaacSim/Labs, ROS2 for sim, control and networking we use for sim. Wanted to know what makes your software special. Would love to be one of the first users.