Articles tagged with "simulation"
World’s fastest supercomputer shows how black holes shape galaxies
Scientists have utilized Frontier, the world’s fastest supercomputer located at Oak Ridge National Laboratory, to simulate how supermassive black holes influence the stability and evolution of galaxy clusters over billions of years. By modeling a black hole with a billion solar masses at the center of a galaxy cluster weighing a quadrillion Suns, researchers tracked the activity of black hole jets and their impact on the surrounding environment. These jets, which move at speeds up to 5% of the speed of light in the simulation, inject heat, dust, and gas into the cluster, regulating energy and preventing the collapse of these massive cosmic structures. The simulation required immense computational resources, including 700,000 node hours and over 17,000 GPUs, highlighting the unique capability of Frontier to handle such large-scale astrophysical problems. The study revealed new insights into the formation of gas filaments around galaxy clusters, phenomena previously observed but never successfully reproduced in simulations. These filaments arise from the turbulence created by interactions between cold gases, hot
energysupercomputerblack-holesastrophysicsgalaxy-clusterssimulationcomputational-scienceJet-powered robot baby shows robust flight control in simulations
The article discusses a significant advancement in flight control for the jet-powered humanoid robot iRonCub, developed by researchers at the Italian Institute of Technology. The team introduced a novel Model Predictive Control (MPC) framework that integrates both the robot’s fast joint actuators and slower jet engine dynamics through a multirate control design. This approach uses a linearized centroidal momentum model for flight dynamics combined with a nonlinear model for jet propulsion, allowing the controller to issue commands at rates suited to each actuator’s physical capabilities. Simulations in MuJoCo demonstrated the robot’s ability to maintain stable flight, recover from disturbances such as pushes causing pitching and rolling, and accurately track both smooth and aggressive flight trajectories. This research marks a step forward in combining walking robots with airborne mobility, with the iRonCub3 platform having recently achieved lift-off in the lab by hovering about 50 centimeters using four jet engines. The unified multirate MPC framework not only improves responsiveness and stability by embedding jet dynamics directly
roboticshumanoid-robotjet-propulsionflight-controlmodel-predictive-controlsimulationautonomous-systemsLQMs vs. LLMs: when AI stops talking and starts calculating
The article discusses the emerging role of Large Quantitative Models (LQMs) as a new class of AI systems that differ fundamentally from Large Language Models (LLMs). Unlike LLMs, which are trained on internet text to generate language-based outputs, LQMs are purpose-built to work with numerical, scientific, and physical data, enabling them to simulate complex real-world systems in fields like chemistry, biology, and physics. Fernando Dominguez, Head of Strategic Partnerships at SandboxAQ—a company at the forefront of AI and quantum technology integration—explains that LQMs can generate novel data not available in existing datasets, such as simulating trillions of molecular interactions. This capability allows LQMs to accelerate drug discovery, financial modeling, and navigation, offering a more quantitative and practical approach to AI-driven innovation. A key example highlighted is SandboxAQ’s collaboration with UCSF’s Institute for Neurodegenerative Diseases, where LQMs enabled the simulation of over 5 million molecular compounds in
materialsAIquantum-computingdrug-discoverysimulationpharmaceuticalscybersecurityNvidia Cosmos Robot Trainer
Nvidia has announced Cosmos, a new simulation and reasoning platform designed to enhance AI, robotics, and autonomous vehicle development. Cosmos aims to enable smarter and faster training of AI models by providing advanced simulation environments that closely mimic real-world scenarios. This approach helps improve the accuracy and efficiency of AI systems used in robotics and autonomous technologies. The platform leverages Nvidia’s expertise in graphics processing and AI to create detailed, realistic simulations that facilitate better decision-making and reasoning capabilities in machines. By accelerating the training process and improving model robustness, Cosmos is expected to advance the development of intelligent robots and autonomous vehicles, ultimately contributing to safer and more reliable AI-driven systems.
robotAINvidiaautonomous-vehiclessimulationrobotics-trainingartificial-intelligenceShengShu Technology launches Vidar multi-view physical AI training model - The Robot Report
ShengShu Technology, a Beijing-based company founded in March 2023 specializing in multimodal large language models, has launched Vidar, a multi-view physical AI training model designed to accelerate robot development. Vidar, which stands for “video diffusion for action reasoning,” leverages a combination of limited physical training data and generative video simulations to train embodied AI models. Unlike traditional methods that rely heavily on costly, hardware-dependent physical data collection or purely simulated environments lacking real-world variability, Vidar creates lifelike multi-view virtual training environments. This approach allows for scalable, robust training of AI agents capable of real-world tasks, reducing the need for extensive physical data by up to 1/80 to 1/1,200 compared to industry-leading models. Built on ShengShu’s flagship video-generation platform Vidu, Vidar employs a modular two-stage learning architecture that separates perceptual understanding from motor control. In the first stage, large-scale general and embodied video data train the perceptual
robotembodied-AIAI-training-modelsimulationgenerative-videorobotics-developmentphysical-AIDeepMind reveals Genie 3, a world model that could be the key to reaching AGI
Google DeepMind has introduced Genie 3, a foundational world model designed as a significant step toward artificial general intelligence (AGI). Unlike previous narrow models, Genie 3 is a real-time, interactive, general-purpose world model capable of generating diverse 3D environments from simple text prompts. It can produce several minutes of photo-realistic or imaginative simulations at 24 frames per second and 720p resolution, surpassing its predecessor’s 10-20 second limit. A key innovation is its ability to maintain physical consistency over time by remembering prior generated content, enabling it to simulate coherent, physically plausible worlds that reflect an intuitive understanding of physics without relying on hard-coded physics engines. Genie 3’s architecture is auto-regressive, generating frames sequentially while referencing earlier frames to maintain continuity and realism. This memory-driven approach allows the model to simulate dynamic scenarios where objects interact naturally, making it an ideal training environment for embodied AI agents. These agents can explore, plan, and learn through trial and
robotartificial-intelligenceworld-modelsimulationembodied-agentsphysics-simulationDeepMindRaindrops at rocket speeds: Water's impact on hypersonic craft revealed
A recent study has revealed how tiny water droplets, such as raindrops, can significantly affect hypersonic vehicles traveling at speeds exceeding Mach 5 (over 6,173 km/h). When these droplets impact a hypersonic aircraft or missile, they tend to break up into smaller droplets that become entrapped and accelerate near the vehicle’s surface. This interaction can disrupt the airflow around the vehicle and increase the likelihood of droplet impacts, especially with larger droplets, potentially affecting the vehicle’s structural integrity. The research team used advanced simulations combining Eulerian and Lagrangian frameworks to model the complex multiphase flow interactions between water droplets and hypersonic airflows. Their findings emphasize the importance of considering droplet breakup dynamics when estimating impact forces on hypersonic vehicles. This work not only aids in the design and development of next-generation hypersonic aircraft but also enhances the fundamental understanding of multiphase flows under extreme conditions. The researchers plan to conduct more detailed simulations to further explore individual
materialshypersonic-vehiclesfluid-dynamicsmultiphase-flowaerospace-engineeringsimulationdroplet-impact99.9% reliable robot vision studio completes week-long task in hours
Apera, a Canadian company, has developed Apera Forge, a web-based, AI-powered 4D vision design studio that significantly accelerates the development of vision-guided robotic (VGR) automation projects. This browser-based platform requires no hardware and enables industrial manufacturers to simulate robotic applications—including parts, grippers, robots, and cell environments—in minutes rather than days. By training AI neural networks through extensive digital cycles, Forge achieves over 99.9% reliability in object recognition and task performance, delivering deployable vision programs within 24 to 48 hours. This drastically reduces the time and risks traditionally involved in creating robotic cells for bin picking, material handling, and de-racking. The latest upgrades to Forge enhance its flexibility and simulation capabilities, supporting advanced robotic cell design with customizable camera placement, bin positioning, and obstacle integration to better replicate real-world conditions. Notably, Forge now supports end-of-arm-tooling (EOAT) mounted camera configurations (Eye-in-Hand), allowing users to
robotAIvision-guided-roboticsautomationindustrial-manufacturingsimulationAI-trainingScientists simulate how tens of thousands of electrons move in real time
materialsenergyquantum-mechanicselectron-dynamicsphotovoltaic-cellssimulationnanostructuresRealtime Robotics launches Resolver for motion planning, simulation
robotmotion-planningautomationindustrial-robotscollision-avoidanceworkcell-designsimulation