Algoryx

Physical AI is a foundational technology for autonomous vehicles and robots, enabling them to interact safely and intelligently with both the real world and human beings. These systems are increasingly powered by foundation models that go beyond today’s ubiquitous Large Language Models (LLMs) to also incorporate vision, language, and action — so-called VLA models. As development accelerates, one emerging challenge is the question of how these models will respond to synthetic training data — particularly data generated by other AI systems. This topic is explored by Algoryx CEO Kenneth Bodin in a recent blog post. As he writes, current VLA models are typically trained on text, large volumes of real-world data such as video and sensor measurements, and physics-based simulations. They are still largely untouched by large-scale use of AI-generated synthetic data — but for how long? Drawing on early findings from LLM research that suggest performance degradation when models are trained on AI-generated content, Kenneth Bodin raises an important question: Could similar recursive effects emerge in models tasked with physical interaction and decision-making? At Algoryx, we see tremendous potential in using science-grounded simulation to train AI systems that are both efficient and reliable — especially when safety and physical realism are non-negotiable. But as Kenneth Bodin writes, more rigorous research is essential to understand and mitigate the risks of synthetic contamination, as "we certainly cannot afford nonsensical robots in healthcare environments or dangerously erratic behavior in autonomous heavy machinery – like 20-ton excavators.” Read the full blog post here: https://xmrrwallet.com/cmx.pwww.linkedin.com/pulse/echoes-nonsense-potential-self-destruction-recursive-ai-kenneth-bodin-z24hf/ And don’t forget to follow Kenneth Bodin for more knowledgeable and thought-provoking takes on physical AI, machine autonomy, and the future of tech. #PhysicalAI #VLAModels #SyntheticData #PhysicsSimulation #TrustworthyAI #MachineAutonomy #AIresearch

Pushing Limits — One Step at a Time This is a nice simulation from Volvo CE, showing how a 6x6 articulated hauler climbs a steep hill by progressively engaging its drivetrain systems. It’s a simple but striking demonstration of how control systems respond in real-world conditions – or in this case, simulated ones. Here’s what happens: Step 1: Engage 6x6 Torque is sent to all six wheels. The machine gains grip — but sometimes, that’s not enough.. Step 2: Lock axle differentials Left and right wheels on each axle are forced to rotate at the same speed, preventing single-wheel spin.. still did not make it Step 3: Connect rear axles All four rear wheels are mechanically synchronized in terms of rotational speed, This prevents unnecessary spinning by ensuring speed equality — not torque equality.. The truck conquers the hill. But that’s just scratching the surface. 🎯 With simulations like this, we can explore much more: - How tipping behavior changes with load and slope - Different engines and drivetrain for vehicles - How autonomy handles path planning, tipping, and avoiding repeated wheel ruts, but also to test and verify AI control systems - How machine flows can be optimized between hauler and loader Simulation is not just about testing — it’s about learning, iterating, and evolving faster. #VolvoCE #Simulation #Autonomy #OffHighway #AGXDynamics #EngineeringExcellence #Innovation

Space Innovation Forum is back in orbit! 17-18 September, we return to Kiruna for two packed days of fresh ideas, bold ventures, and real conversations about where innovation is headed and possibilities occur. From student-driven pitches and business insights to deep dives into sustainability and space tech, this year’s forum is all about reconnecting and reimagining. It’s a space to share what has worked, what is next, and what is possible – together. 🎤 Speakers include: ✅ Ernesto Gutiérrez, Rymdstyrelsen ✅ Anne Lidgard, Vinnova ✅ Olle Norberg, Swedish Institute of Space Physics ✅ Sandra Nilsson, Arctic Space ✅ Rene Laufer, Luleå University of Technology ✅ Johanna Bergstrom-Roos Space Sweden North/LTU Business ✅ Kenneth Bodin, Algoryx ✅ Philip Påhlsson, EISCAT ✅ Liisa Kiviloog Emilson SSC - Swedish Space Corporation ✅ Per Bodin, OHB Sweden AB ✅ Léo Ewen Trapier, SSC - Swedish Space Corporation ✅ Wilhelm Dingertz, ECAPS ✅ Abhilash Vakkada Ramachandran, Ph.D., Hydromars ✅ Mats Tyni, SSC - Swedish Space Corporation ✅ Karoline Pettersson, LTU Business ✅ Jaan Praks, Aalto University …and more 🛰 Key topics: → Open innovation & foresight: what suppliers will the space sector need tomorrow? → Access to space & ecosystem functions to boost space → Payload Drop: Student Innovation Pitches → Sustainability — what Suppliers Need to Know About Sustainability in the Space Supply Chain 📍 Where: Kiruna, Sweden 📅 When: September 17–18, 2024 🎟️ Price: Free (but space is limited — no-show fees apply) 🔗 Register now → https://xmrrwallet.com/cmx.plnkd.in/dvbdp3Y6 💡 This is your launchpad — whether you're a startup, a student, or someone who believes that the next big leap starts here, on Earth. Organizers and sponsors: This event is organized through Space Sweden North - an innovation partnership between Luleå University of Technology, LTU Business, Swedish Institute of Space Physics, SSC - Swedish Space Corporation Arctic Business and Aerospace Cluster Sweden (ACS). Space Sweden North is co-funded by the European Regional Development Fund, Utveckla Norrbotten - en del av Region Norrbotten, Kiruna kommun and Sparbanken Nord. #SpaceInnovationForum #SpaceNorth #SustainabilityInSpace #Kiruna2024 #NewSpace #InnovationEcosystem #ArcticTestbed #UpstreamSpace #DiversityInTech

Track your mining equipment in real time, save fuel, and minimize safety issues. That’s modern mining with simulation-first approaches and autonomous solutions offered by SoftServe and Algoryx. Download the white paper: https://xmrrwallet.com/cmx.psftsrv.com/z4BwpK

Building a complete virtual test environment for mining automation is no small task. Dynamic machine models, control systems, virtual sensors, and simulated environments must integrate seamlessly and function together in real-time. Achieving this integration requires significant effort and expertise. An engineering team from SoftServe has tackled this challenge head-on in a new white paper and demo, constructing a sophisticated virtual mining site featuring realistic vehicles, terrain, and material behaviors designed to support the testing and development of autonomous systems. At the core of Softserve’s solution architecture is AGX Dynamics from Algoryx, delivering real-time, engineering-grade physics to capture the critical dynamics of excavation, wheel-soil interaction, and fuel consumption. Key capabilities demonstrated in the white paper include ▪️ Multi-robot navigation ▪️ Excavation and dumping ▪️ Material handling ▪️ Terrain interaction modeling ▪️ Hybrid-electric and combustion engine simulation ▪️ #ROS2-based vehicle control and multi-modal #SLAM A real-life mining site was fully recreated in #Unity using photogrammetry and point cloud data, with AGX Dynamics for Unity providing high-fidelity simulations of machine-environment interactions. Softserve’s work shows how a simulation-first approach can accelerate mining automation, providing a platform for safe, repeatable, and scalable development of autonomous behaviors. 📄 Download the white paper, "Dig Deeper in Mining with Automation and Simulation," to explore how these technologies come together in practice: https://xmrrwallet.com/cmx.plnkd.in/dCcP3znk

We’re excited to team up with Algoryx in the XSCAVE project! Known for their powerful AGX Dynamics engine, Algoryx is a leader in real-time physics simulation—think accurate modeling of friction, hydraulics, and deformable terrain. It’s tech that makes heavy off-road machines smarter and safer. In XSCAVE, Algoryx is helping push the boundaries by embedding a lightweight physics engine directly into machines like excavators and forestry forwarders 🌲🚧. This lets the robots simulate and reason about their actions in real time—like predicting how terrain will respond before making a move. That’s a big deal when operating in rough, unpredictable environments 🏔️.

Optimized routing can lead to up to 15% energy savings and a 15% longer equipment lifespan. SoftServe and Algoryx make this happen with robotics, data analytics, and high-fidelity simulations. Read the article: https://xmrrwallet.com/cmx.psftsrv.com/4ywCmB

Algoryx is entering the domain of humanoids and nextgen robotics. We simulate full coordinate representations of compliant articulated legged robots with accurate frictional contacts, dynamic terrain and all the things you can simulate in AGX Dynamics. We also provide a full fledged gymnasium environment for deep reinforcement learning, as well as inverse dynamics, inverse kinematics, ROS support, sensor simulations and more. 3D in our default environment, or with Unity or Unreal Engine. Most other approaches, even from trillion dollar bigtech, fall back on simplified dynamics, with reduced coordinates for idealized robots and over simplified interaction. This doesn't hold if you want your robots to be competitive, fast, dexterious,stable and energy efficient with highest possible sim-2-real for advanced actions. As always, we do the real thing at Algoryx! High fidelity, faster than realtime rich multidomain physics with massive parallelism for the synthetic data generation you need for machine learning. We don't have alot of access to hardware yet so I'd be glad to get more collaboration on this! Omar has explored humanoid walking cycles on varying terrain conditions in his MSc project.

Our Silver Sponsors Algoryx were on stage next with their keynote on ‘Simulation for the Development of Safe, Smart Machines’ delivered by their Senior Business Director, Hugo Börjesson and Senior Business Developer, Christer Ekholm. The pair spoke about testing the untestable by using simulation to explore hazardous, rare, and extreme scenarios that are unsafe or impossible to test physically. They shared how you are able to see what machines see with realistic sensor simulation that is critical for verifying perception, decision-making, and safe autonomous behaviour. They also gave their insights on validating the software-defined vehicle with continuous test and verification of the control/autonomy stack. #Algoryx #Innovatrix #AOMTEUROPE #Berlin #Construction #AOMT #ADAS #Connectivity #GNSS #Navigation #IOT #AutonomousVehicles