A new AI model replaces months of simulation with near-instant predictions, changing how spacecraft operations are prepared
Updated
April 24, 2026 10:53 AM
Northrop Grumman Stargaze serves as the mother ship for the Pegasus, an air-launched orbital rocket. PHOTO: UNSPLASH
Flexcompute, a startup that builds software to simulate real-world physics, is working with Northrop Grumman to change how space missions are prepared. Together, they have developed an AI-based system that can predict how spacecraft respond during critical manoeuvres such as docking—when one spacecraft moves in and connects with another in orbit. These steps have traditionally taken months of preparation.
At the centre of this work is a long-standing problem in space operations. When a spacecraft fires its thrusters, the exhaust plume interacts with nearby surfaces. These interactions can affect movement, temperature and stability. Because these effects are difficult to test in real conditions, engineers have relied on large volumes of computer simulations to estimate outcomes before a mission. That process is slow and resource-intensive.
The new system replaces much of that workflow with a trained AI model. Instead of running millions of simulations, the model learns patterns from physics-based data and can make predictions in seconds. It also provides a measure of uncertainty, which helps engineers understand how reliable those predictions are when making decisions.
"At Northrop Grumman, we're pioneering physics AI to accelerate design and solve complex simulation and modelling problems like plume impingement—critical for station keeping, rendezvous and space robotics. Simply put: we're pushing the boundaries of advanced space operations", said Fahad Khan, Director of AI Foundations at Northrop Grumman. "Partnering with Flexcompute and NVIDIA, we're accelerating innovation and mission timelines to deliver superior space capabilities for customers at the speed they need".
The system is built using technology from NVIDIA, which provides the computing framework behind the model. Flexcompute has adapted it to handle the specific challenges of spaceflight, including how gases expand and interact in a vacuum. The result is a tool that can simulate complex scenarios much faster while maintaining the level of accuracy needed for mission planning.
By shortening preparation time, the model changes how engineers approach spacecraft design and operations. Faster predictions mean teams can test more scenarios and adjust plans more quickly. It also helps improve fuel use and extend the lifespan of spacecraft.
"Northrop Grumman's confidence reflects what sets Flexcompute apart", said Vera Yang, President and Co-Founder of Flexcompute. "We are able to take the most accurate and scalable physics foundations and evolve them into highly trained, customized Physics AI solutions that engineers can rely on. This work shows how we are transforming the role of simulation, not just speeding it up, but expanding what engineers can confidently solve and how quickly they can act".
The collaboration points to a broader shift in how engineering problems are being handled. Instead of relying only on detailed simulations that take time to run, companies are beginning to use AI systems that can approximate those results quickly while still reflecting the underlying physics.
"The industry's most ambitious space missions now demand a level of speed and precision that traditional engineering cycles can no longer sustain", said Tim Costa, vice president and general manager of computational engineering at NVIDIA. "By integrating NVIDIA PhysicsNeMo, Northrop Grumman and Flexcompute are transforming complex simulations like plume impingement from days of compute into seconds of insight, drastically accelerating the path from mission concept to orbit".
What emerges from this work is a shift in how missions are prepared. When prediction cycles move from months to seconds, testing and decision-making can happen faster. For space operations, where timing and precision are closely linked, that change could reshape how systems are built and run.
Keep Reading
Robots that learn on the job: AgiBot tests reinforcement learning in real-world manufacturing.
Updated
January 8, 2026 6:34 PM
A humanoid robot works on a factory line, showcasing advanced automation in real-world production. PHOTO: AGIBOT
Shanghai-based robotics firm AgiBot has taken a major step toward bringing artificial intelligence into real manufacturing. The company announced that its Real-World Reinforcement Learning (RW-RL) system has been successfully deployed on a pilot production line run in partnership with Longcheer Technology. It marks one of the first real applications of reinforcement learning in industrial robotics.
The project represents a key shift in factory automation. For years, precision manufacturing has relied on rigid setups: robots that need custom fixtures, intricate programming and long calibration cycles. Even newer systems combining vision and force control often struggle with slow deployment and complex maintenance. AgiBot’s system aims to change that by letting robots learn and adapt on the job, reducing the need for extensive tuning or manual reconfiguration.
The RW-RL setup allows a robot to pick up new tasks within minutes rather than weeks. Once trained, the system can automatically adjust to variations, such as changes in part placement or size tolerance, maintaining steady performance throughout long operations. When production lines switch models or products, only minor hardware tweaks are needed. This flexibility could significantly cut downtime and setup costs in industries where rapid product turnover is common.
The system’s main strengths lie in faster deployment, high adaptability and easier reconfiguration. In practice, robots can be retrained quickly for new tasks without needing new fixtures or tools — a long-standing obstacle in consumer electronics production. The platform also works reliably across different factory layouts, showing potential for broader use in complex or varied manufacturing environments.
Beyond its technical claims, the milestone demonstrates a deeper convergence between algorithmic intelligence and mechanical motion.Instead of being tested only in the lab, AgiBot’s system was tried in real factory settings, showing it can perform reliably outside research conditions.
This progress builds on years of reinforcement learning research, which has gradually pushed AI toward greater stability and real-world usability. AgiBot’s Chief Scientist Dr. Jianlan Luo and his team have been at the forefront of that effort, refining algorithms capable of reliable performance on physical machines. Their work now underpins a production-ready platform that blends adaptive learning with precision motion control — turning what was once a research goal into a working industrial solution.
Looking forward, the two companies plan to extend the approach to other manufacturing areas, including consumer electronics and automotive components. They also aim to develop modular robot systems that can integrate smoothly with existing production setups.