Engineering Simulation Tools Overview

Exciting updates on Project GR00T! We discover a systematic way to scale up robot data, tackling the most painful pain point in robotics. The idea is simple: human collects demonstration on a real robot, and we multiply that data 1000x or more in simulation. Let’s break it down: 1. We use Apple Vision Pro (yes!!) to give the human operator first person control of the humanoid. Vision Pro parses human hand pose and retargets the motion to the robot hand, all in real time. From the human’s point of view, they are immersed in another body like the Avatar. Teleoperation is slow and time-consuming, but we can afford to collect a small amount of data.  2. We use RoboCasa, a generative simulation framework, to multiply the demonstration data by varying the visual appearance and layout of the environment. In Jensen’s keynote video below, the humanoid is now placing the cup in hundreds of kitchens with a huge diversity of textures, furniture, and object placement. We only have 1 physical kitchen at the GEAR Lab in NVIDIA HQ, but we can conjure up infinite ones in simulation. 3. Finally, we apply MimicGen, a technique to multiply the above data even more by varying the *motion* of the robot. MimicGen generates vast number of new action trajectories based on the original human data, and filters out failed ones (e.g. those that drop the cup) to form a much larger dataset. To sum up, given 1 human trajectory with Vision Pro  -> RoboCasa produces N (varying visuals)  -> MimicGen further augments to NxM (varying motions). This is the way to trade compute for expensive human data by GPU-accelerated simulation. A while ago, I mentioned that teleoperation is fundamentally not scalable, because we are always limited by 24 hrs/robot/day in the world of atoms. Our new GR00T synthetic data pipeline breaks this barrier in the world of bits. Scaling has been so much fun for LLMs, and it's finally our turn to have fun in robotics! We are creating tools to enable everyone in the ecosystem to scale up with us: - RoboCasa: our generative simulation framework (Yuke Zhu). It's fully open-source! Here you go: http://robocasa.ai - MimicGen: our generative action framework (Ajay Mandlekar). The code is open-source for robot arms, but we will have another version for humanoid and 5-finger hands: https://lnkd.in/gsRArQXy - We are building a state-of-the-art Apple Vision Pro -> humanoid robot "Avatar" stack. Xiaolong Wang group’s open-source libraries laid the foundation: https://lnkd.in/gUYye7yt - Watch Jensen's keynote yesterday. He cannot hide his excitement about Project GR00T and robot foundation models! https://lnkd.in/g3hZteCG Finally, GEAR lab is hiring! We want the best roboticists in the world to join us on this moon-landing mission to solve physical AGI: https://lnkd.in/gTancpNK

🔌 Grid operators are implementing various strategies to manage the declining inertia caused by the increased penetration of variable generation (VG) resources, such as wind and solar. These strategies fall into three main categories: maintaining inertia, providing more response time, and enhancing fast frequency response. To maintain inertia, operators can ensure that a mix of synchronous generators is online to exceed critical inertia levels. Additionally, synchronous renewable energy sources and synchronous condensers can be deployed to provide inertia. To provide more response time, operators can reduce contingency sizes and adjust underfrequency load shedding (UFLS) settings. Finally, enhancing fast frequency response involves leveraging load resources, extracting wind kinetic energy, and dispatching inverter-based resources to improve the grid's ability to respond to frequency changes. 🍃 Extracted wind kinetic energy refers to the capability of wind turbines to provide fast frequency response (FFR) by utilising the kinetic energy stored in their rotating blades. This approach can be particularly effective in addressing the challenges posed by declining inertia in power systems with high wind penetration. By extracting kinetic energy, wind turbines can respond rapidly to frequency deviations, thereby helping to stabilise the grid. This method can be used in conjunction with other resources to enhance overall system reliability and maintain frequency within acceptable limits. 💡 High deployment of variable generation (VG) resources can be effectively managed by combining extracted kinetic energy from wind turbines and increasing output from curtailed wind plants. The figure below illustrates that when these two strategies are combined, they significantly mitigate frequency decline. The simulation shows that relying solely on extracted kinetic energy results in frequency falling below UFLS (underfrequency load shedding), while using only FFR barely avoids UFLS. However, when both methods are applied together, the frequency decline is minimal, demonstrating that these approaches can serve as viable alternatives to traditional inertia and primary frequency response from conventional generators. #gridmodernization #stability #gridforming #powerelectronics #renewables #cleanenergy #solidstate

A critical challenge in modern grid stability is that inverter-based resources (IBRs) are often “black boxes” to utilities and system operators. Inverter manufacturers and plant developers understandably hesitate to disclose proprietary control strategies, leaving operators with limited visibility into internal dynamics. The problem is further compounded by the fact that IBRs can switch among multiple control modes, which are typically unknown to operators yet can exhibit dramatically different dynamic behaviors. In the final days of 2025, we were excited to learn that our paper on black-box IBR modeling was accepted by IEEE Transactions on Smart Grid. In this work, we develop a comprehensive data-driven framework that uses only terminal measurements to discover unknown control modes and learn continuous-time models that accurately capture IBR dynamics under each mode. By leveraging physics-inspired deep learning, the proposed approach addresses four major challenges in a unified way: 🚀 High-Order Nonlinear Representation Using only terminal measurements, the framework provides a general learning approach for characterizing arbitrary high-order nonlinear dynamics of IBRs. It is not tied to any specific control paradigm and can cover anything from power/voltage/current control loops to virtual synchronous machines (VSMs) and phase-locked loops (PLLs). 🚀 Continuous-Time Modeling Unlike most data-driven methods built on discrete-time models (e.g., RNNs, LSTMs, Transformers), our approach learns continuous-time state-space models (differential-algebraic equations). This enables seamless integration of the learned IBR models into standard power-system time-domain simulations with arbitrary numerical integration schemes and step sizes. 🚀 Discovery of Unknown Control Modes A physics-inspired deep unsupervised learning mechanism automatically identifies distinct control modes from historical disturbance data and learns separate state-space models that represent the dynamics associated with each mode. 🚀 Robustness to Noise and Uncertainty Inspired by Kalman filtering, the learning architecture explicitly accounts for system uncertainties and measurement noise, both of which are ubiquitous in real-world grid systems and data. It ensures the method’s robust performance in practical settings. The examples in the paper demonstrate how the proposed framework can learn accurate time-domain models of fully black-box IBRs and deliver highly accurate long-horizon predictions of their responses to grid disturbances, e.g., subsynchronous oscillations caused by PLL interactions in weak grids. See details here: https://lnkd.in/eFd5CU4e #PowerSystem #SmartGrid #InverterBasedResources #RenewableEnergy #PowerElectronics #Control #PowerSystemStability #PowerSystemModeling #PowerSystemSimulation #SystemIdentification #DataDriven #MachineLearning #DeepLearning #ArtificialIntelligence #PhysicsInformed #IEEETransactionsOnSmartGrid

As we have all been saying, the grid is no longer just an engineering challenge—it’s the primary bottleneck for the future of AI and the energy transition. The barrier: human and bureaucratic processes. Who has a solution for this? At CERAWeek 2026 this week, the atmosphere has shifted from "How do we decarbonize?" to a much more urgent "How do we plug in?" With interconnection queues stretching 5–10 years and turbine lead times hitting 2030, speed to power is the new global currency. A new wave of "Grid-Tech" companies is moving past legacy manual processes to solve the bottleneck through software, digital twins, and flexible load. Here are the innovators leading the charge to break the logjam: 1. As I wrote in my last post, NVIDIA & Emerald AI’s solution: By treating AI data centers as "virtual batteries," this software allows hyperscalers to bypass years of grid study. Instead of a fixed-load connection, they use AI to dynamically flex power consumption during grid stress. This "flexible interconnection" model could unlock up to 100 GW of capacity by optimizing the grid we already have. 2. Enverus (Pearl Street Technologies)’s solution: Interconnect™ (Study Automation) The manual process of "power flow studies" is a primary cause of queue delays. Enverus is using its SUGAR™ engine to automate these complex reliability simulations, reducing the time required for interconnection studies from months to just a few days. 3. @Tapestry (X, The Moonshot Factory)’s solution: Grid Digital Twin (Visibility) I’ve been excited about Tapestry building a high-fidelity "Google Maps for electrons." By creating a unified digital twin of the grid, they allow operators like PJM to run transient simulations in real-time, identifying exactly where new projects can fit without triggering expensive, time-consuming network upgrades. 4. Neara The Solution: 3D Infrastructure Modeling (Reconductoring) Before building new towers, we must maximize existing ones. Neara’s platform uses 3D digital twins to simulate "reconductoring"—replacing old wires with high-capacity advanced conductors. This allows developers to find "low-hanging fruit" capacity that can be brought online in a fraction of the time. 5. GridStatus The Solution: Real-Time Data Transparency You can't manage what you can't see. GridStatus has become the de facto data layer for the energy transition, providing the real-time transparency into grid congestion and pricing that developers need to site projects where the grid can actually handle them. The technology is ready. The capital is waiting. We need regulatory frameworks to keep pace with these digital solutions. #CERAWeek #CleanTech #EnergyTransition #GridModernization #AI #DataCenters #SpeedToPower

A few years ago, I learned the hard way that jumping straight into hardware, sensors, motors, and wiring can lead to costly mistakes and late-night headaches. That’s when I discovered the true importance of #simulation in robotics and engineering. During the early phase of my final-year thesis, I spent weeks recreating our school cafeteria with Iman Tokosi in Blender, exporting it as an SDF model and loading it into Gazebo using #ROS2. Suddenly, I could drive a virtual robot through aisles and around tables without the fear of damaging anything real. It was challenging and eye-opening, and it saved me countless hours and resources. Then came the moment that changed everything: integrating #SLAM so the robot could build its own map while moving, and setting up #Nav2 to let it plan and follow paths autonomously. Watching it navigate the environment with precision and independence was a powerful confirmation that the system worked. Now, imagine a world where every structure, product, and system is simulated down to the smallest detail. The result? Reduced costs, faster development, increased reliability, enhanced safety, and stronger adherence to standards. Some may still view simulation as “just for show,” but I’ve experienced firsthand that it’s the foundation of true innovation. Are you leveraging simulation in your next robotics or engineering project? Let’s connect and exchange ideas!

You cannot train physical AI on reality alone. There is not enough of it. Jensen Huang explains why NVIDIA built Alpamayo, a robotics model that learns from synthetic data grounded in physics. The problem is fundamental. Teaching physical AI like autonomous vehicles or robotics requires vast amounts of diverse interaction data. Videos exist. Lots of videos. But hardly enough to capture the diversity and type of interactions needed. So NVIDIA transformed compute into data. Using synthetic data generation grounded and conditioned by laws of physics, they can selectively generate training scenarios reality cannot provide. The example Huang shows is remarkable. A basic traffic simulator output gets fed into Cosmos AI world model. What emerges is physically based, physically plausible surround video that AI can learn from. This solves a constraint that limited physical AI development. You cannot train autonomous systems on every possible scenario by recording reality. There are not enough cameras, time, or situations. But you can simulate physics accurately enough that AI trained on synthetic data generalizes to real environments. Why this matters beyond autonomous vehicles. Any AI learning physical interactions faces the same data scarcity problem. Manufacturing robots, warehouse automation, infrastructure inspection, medical robotics. All require training on scenarios that are rare, dangerous, or impossible to capture at scale. Synthetic data generation grounded in physics laws becomes essential infrastructure for physical AI deployment. The organizations building AI for physical systems will either master synthetic data generation or remain limited by whatever reality they can record. Watch the full presentation to hear Huang explain how Alpamayo generates training data for autonomous vehicles that think like humans. What physical AI application needs synthetic data because reality cannot provide enough examples?

I had the chance a few weeks ago to sit down (virtually of course) with Woodrow Bellamy III, host of SAE’s Aerospace & Defense Technology podcast, for a candid conversation about virtual prototyping. While I work across many industries these days, I always enjoy a chance to share insights from my years working in Aerospace. The key question was about the reliability of virtual prototyping as a replacement for physical prototyping. How much faith can be put into a digital model for accurately simulating a real-word system, when you have real-world stakes and consequences? The answer is: a lot. When we first started using simulation for some of the aircraft programs I worked, we questioned if we could trust the results to predict peformance of a new aircraft. It was an appropriate question to ask. It was only after conducting extensive testing to validate the results of the simulation against an existing aircraft that we started to trust the simulation models while designing a new aircraft. The same litmus test will be needed for companies in any industry to take the leap with virtual prototyping: A company can start by developing the virtual model and ensure the validity and robustness of the simulation by comparing against existing physical products. Once the team is sufficiently confident about the performance of the digital model, they can take the insights from existing products and systems and apply them toward the development of future projects. The digital twin of a physical system is infinitely more malleable. It can be designed, tested, and experimented upon with far more ease and using significantly fewer resources. Virtual prototypes allow for limitless design exploration, help to identify design issues early and before building physical prototypes, and make physical testing more effective. Before going for the real-life testing, virtual analyses highlight critical areas in the design, and the test plans can be adjusted to focus on areas of greatest concern. The aerospace industry adopted the use of the digital twin as a revolutionary design tool decades ago. And the digital twin has evolved quite a bit over the ensuing decades. It is no longer just a 3D model of a product or process. Today, the comprehensive digital twin is a precise virtual representation of the product that replicates its physical form, function and behavior and encompasses all cross-domain models and data from mechanical and electrical through software code. So, although the current ratio of prototyping testing, worldwide, may be 90% physical and 10% digital, it isn’t an overstatement to conceive of a future where the ratio is flipped; maybe even 100% digital. I'm excited about the opportunities offered by virtual prototyping and testing as a means to enable companies to develop and validate innovative products faster! To check out my conversation with Woodrow, please check out the link in the comments. #digitaltransformation #siemensxcelerator

𝗗𝗼𝗻’𝘁 𝗝𝘂𝘀𝘁 𝗥𝗲𝗮𝗱 𝗔𝗯𝗼𝘂𝘁 𝗔𝗜 𝗶𝗻 𝗠𝗮𝗻𝘂𝗳𝗮𝗰𝘁𝘂𝗿𝗶𝗻𝗴. 𝗔𝗽𝗽𝗹𝘆 𝗜𝘁. The AI headlines are exciting. But if you're a founder, engineer, or educator in manufacturing, here's the question that actually matters: 𝗪𝗵𝗮𝘁 𝗰𝗮𝗻 𝘆𝗼𝘂 𝗱𝗼 𝘵𝘰𝘥𝘢𝘺 𝘁𝗼 𝘁𝘂𝗿𝗻 𝘁𝗵𝗲𝘀𝗲 𝗶𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻𝘀 𝗶𝗻𝘁𝗼 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻? Let’s get tactical. 𝟭. 𝗦𝘁𝗮𝗿𝘁 𝘄𝗶𝘁𝗵 𝗔𝗜 𝗱𝗲𝗺𝗮𝗻𝗱 𝗳𝗼𝗿𝗲𝗰𝗮𝘀𝘁𝗶𝗻𝗴 Tool to try: Lenovo’s LeForecast A foundation model for time-series forecasting. Trained on manufacturing-specific datasets. 𝗨𝘀𝗲 𝗶𝘁 𝗶𝗳: You’re battling supply chain volatility and need better inventory planning. 👉 Tip: Start by connecting your ERP data. Don’t wait for perfect integration: small wins snowball. 𝟮. 𝗕𝘂𝗶𝗹𝗱 𝗮 𝗱𝗶𝗴𝗶𝘁𝗮𝗹 𝘁𝘄𝗶𝗻 𝗯𝗲𝗳𝗼𝗿𝗲 𝗯𝘂𝘆𝗶𝗻𝗴 𝘁𝗵𝗮𝘁 𝗻𝗲𝘅𝘁 𝗿𝗼𝗯𝗼𝘁 Tools behind the scenes: NVIDIA Omniverse, Microsoft Azure Digital Twins Schaeffler + Accenture used these to simulate humanoid robots (like Agility’s Digit) inside full-scale virtual factories. 𝗨𝘀𝗲 𝗶𝘁 𝗶𝗳: You’re considering automation but can’t afford to mess up your live floor. 👉 Tip: Simulate your current workflows first. Even without a robot, you’ll find inefficiencies you didn’t know existed. 𝟯. 𝗕𝗿𝗶𝗻𝗴 𝘆𝗼𝘂𝗿 𝗤𝗔 𝗽𝗿𝗼𝗰𝗲𝘀𝘀 𝗶𝗻𝘁𝗼 𝘁𝗵𝗲 𝟮𝟬𝟮𝟬𝘀 Example: GM uses AI to scan weld quality, detect microcracks, and spot battery defects: before they become recalls. 𝗨𝘀𝗲 𝗶𝘁 𝗶𝗳: You’re relying on spot checks or human-only inspections. 👉 Tip: Start with one defect type. Use computer vision (CV) models trained with edge devices like NVIDIA Jetson or AWS Panorama. 𝟰. 𝗘𝗱𝗴𝗲 𝗶𝘀 𝗻𝗼𝘁 𝗼𝗽𝘁𝗶𝗼𝗻𝗮𝗹 𝗮𝗻𝘆𝗺𝗼𝗿𝗲 Why it matters: If your AI system reacts in seconds instead of milliseconds, it's too late for safety-critical tasks. 𝗨𝘀𝗲 𝗶𝘁 𝗶𝗳: You're in high-speed assembly lines, robotics, or anything safety-regulated. 👉 Tip: Evaluate edge-ready AI platforms like Lenovo ThinkEdge or Honeywell’s new containerized UOC systems. 𝟱. 𝗕𝗲 𝗲𝗮𝗿𝗹𝘆 𝗼𝗻 𝗰𝗼𝗺𝗽𝗹𝗶𝗮𝗻𝗰𝗲 The EU AI Act is live. China is doubling down on "self-reliant AI." The U.S.? Deregulating. 𝗨𝘀𝗲 𝗶𝘁 𝗶𝗳: You're deploying GenAI, predictive models, or automation tools across borders. 👉 Tip: Start tagging your AI systems by risk level. This will save you time (and fines) later. Here are 5 actionable moves manufacturers can make today to level up with AI: pulled straight from the trenches of Hannover Messe, GM's plant floor, and what we’re building at DigiFab.ai. ✅ Forecast with tools like LeForecast ✅ Simulate before automating with digital twins ✅ Bring AI into your QA pipeline ✅ Push intelligence to the edge ✅ Get ahead of compliance rules (especially if you operate globally) 🧠 Each of these is something you can pilot now: not next quarter. Happy to share what’s worked (and what hasn’t). 👇 Save and repost. #AI #Manufacturing #DigitalTwins #EdgeAI #IndustrialAI #DigiFabAI

Closing the sim-to-real gap in humanoid robotics requires massive simulation throughput and high-fidelity physics validation. WPP recently detailed their engineering pipeline, showing how they reduced reinforcement learning cycle times for complex humanoid locomotion from 24 hours down to less than 60 minutes. The hardware architecture relies on Google Cloud’s new G4 VMs (powered by NVIDIA RTX PRO 6000 Blackwell GPUs) running NVIDIA Isaac Sim, integrated closely with DeepMind’s MuJoCo physics engine. The mechanics: The team mapped raw human mocap data (over 200 degrees of freedom) down to a constrained 29-DOF OpenUSD digital twin. By leveraging a P2P GPU topology to bypass central processing bottlenecks, the infrastructure executed over 3 billion simulations in under an hour. The virtual environment continuously introduced physical micro-variances—simulated pushes, shifting floor friction, and momentum changes—to train the model against the chaos of the real world. The resulting reinforcement learning model was condensed into a highly efficient ONNX policy and deployed directly to the physical robot. This edge policy processes live IMU and joint telemetry to output immediate, stabilized motor commands. Reaching this scale of simulation volume is the precise engineering mechanism that allows control policies to handle unstructured physical deployment. To support the research, Unitree has open-sourced the underlying RL code on GitHub. Blog post : https://lnkd.in/g4-gWzTP #Robotics #PhysicalAI #ReinforcementLearning #MuJoCo #GoogleCloud #IsaacSim #Engineering

With the growing integration of renewable energy, grid stability and reliability are more important than ever. To tackle these challenges, I developed and tested a grid-forming inverter model integrated with droop control, using MATLAB Simulink. 🔌 Key Design Parameters In my design: Injected Power: 10 kW Grid Voltage: 400 V RMS Grid Frequency: 50 Hz This setup represents a real-world scenario where the grid operates slightly off-nominal frequency, and the inverter must dynamically regulate power output to stabilize the system. ⚙️ Why Droop Control? Droop control emulates the behavior of conventional generators, allowing: 1️⃣ Frequency Regulation: Adjusting active power to counter frequency deviations. 2️⃣ Voltage Stability: Sharing reactive power to maintain voltage levels. 3️⃣ Scalability: Supporting multiple inverters without communication. 📊 Simulation Highlights Using MATLAB Simulink, I modeled and simulated the inverter's performance under various grid conditions. Key outcomes include: Stable Frequency Response: The inverter successfully brought the frequency closer to nominal under dynamic loads. Reactive Power Sharing: Demonstrated effective voltage control between parallel inverters. Grid Resilience: Maintained stability during grid disturbances, showcasing the robustness of the droop control approach. 🔍 Insights from the Simulation The simulation confirmed that grid-forming inverters can ensure system stability even in challenging grid conditions, paving the way for integrating higher shares of renewable energy. #MATLAB #SIMULINK #Gridforming #droopcontrol #renewables #PV