Artificial intelligence is evolving at an unprecedented pace, advancing from simple generative tasks to autonomous decision-making through agentic models. Now AI is moving beyond the digital realm into the physical world. This next frontier, known as physical AI, combines advanced algorithms with sensors and actuators, enabling machines to perceive, reason, and act in complex real-world environments.
At Hexagon LIVE 2025 in Las Vegas, a physical AI humanoid robot named AEON made its debut. Think Tron meets I, Robot, but powered by next-gen AI and hardware rather than movie magic.
Developed by Sweden-based industrial tech giant Hexagon in partnership with Nvidia, AEON is designed for real-world industrial work. It can inspect equipment in cramped industrial corridors, navigate hazardous construction zones, and manage logistics in understaffed warehouses.
Spencer Huang, Product Lead for Robotics at Nvidia and the son of CEO Jensen Huang, sees a massive opportunity in humanoids, as their form allows them to perform tasks that are dangerous and demanding for humans.
“Humanoids are one of the many embodiments of Physical AI. To be able to perform human-like requires powerful brains trained on massive amounts of data,” Huang tells Fast Company. “During the test phase, the biggest challenge was to ensure these robots can safely perform tasks in complex, dynamic environments.”
Humanoid robots like AEON aim to help solve an emerging labor crisis. A report from the World Economic Forum shows nearly 50 million manufacturing and logistics jobs remain unfilled globally. In the U.S. alone, the manufacturing sector will need up to 3.8 million new workers by 2033. Hexagon plans to deploy AEON across key industrial applications, including sorting and moving parts, inspecting for defects and compliance, and performing complex tasks such as precision scanning with high-end sensors.
What sets AEON apart from other humanoids, including Tesla’s Optimus Gen 2 and Figure AI’s Helix, is its ability to learn. Traditionally, training industrial robots requires months of manual programming and physical testing. AEON skips that. Tasks such as balance, locomotion, and precision manipulation, which usually take five to six months of coding and trial-and-error, took AEON just two to three weeks.
The key lies in AEON’s ability to generate its own synthetic training data, learn through autonomous simulation and reinforcement, and apply those lessons in the real world. This self-learning AI loop marks a shift in how quickly physical AI systems can adapt to complex environments.
“Simulation helps solve the robotics data challenge, enabling faster, safer development and testing cycles. It also cuts costs by reducing the need for physical prototypes and hardware,” Huang says. “A simulation-first approach lays the groundwork for robots to ultimately improve on their own and even generate new scenarios to challenge themselves, all in virtual environments.”
A Platform Shift for Physical Intelligence
To build AEON, Hexagon used Nvidia’s AI supercomputers to train and fine-tune foundation models; the Nvidia Omniverse platform to test and optimize those models in simulation; and IGX Thor robotic computers to run the models on the robot itself. The company also employed Isaac Sim, a robotic simulation tool built on Omniverse, to train AEON on tasks.
AEON features 22 multimodal sensors and 12 cameras that enable AI-based spatial awareness, asset scanning, and digital twin creation, without needing retraining for each new environment.
“AEON’s wheel-based locomotion allows it to traverse factories and pivot in all directions at speed,” says Arnaud Robert, President of Hexagon’s Robotics division. “It has a battery self-swap mechanism by which it can change its own battery with no downtime, allowing for continuous operations. These design choices separate it from what we have seen on the market.”
For initial training, human demonstrations are collected using teleoperation tools such as the Apple Vision Pro, which streams natural hand movements into a simulated environment in Isaac Lab.
“These authentic human actions serve as the foundation for all subsequent data generation. Synthetic motion generation takes these demonstrations and creates a large number of new motion trajectories,” says Huang. “This multiplies the available training data, enabling robots to learn from a much wider range of scenarios than manual collection alone.”
Hexagon is also exploring the use of Nvidia’s Isaac GR00T N1.5 open foundation model to enhance AEON’s reasoning capabilities, and GR00T-Mimic to generate larger volumes of synthetic motion data from just a few human demonstrations.
Unlike many humanoids still in the R&D phase, AEON is headed for production. It is already set to pilot in real-world environments with German automotive company Schaeffler and Swiss aircraft manufacturer Pilatus, taking on tasks ranging from part inspection to reality capture.
“AEON can do a reality capture of an average factory in an hour, and this could be done multiple times a day,” says Robert. “AEON’s awareness and spatial intelligence algorithms give us a significant edge. If the humanoid is moving quickly across a factory floor and detects a person 10 feet away, it will automatically slow down or adjust its trajectory to avoid getting too close.”
A New Playbook for Robotics
AEON suggests that physical AI is catching up fast. Could its simulation-first, agentic approach be the blueprint for a new class of physical AI systems?
“Nearly every, if not all, humanoid company is exploring and embracing simulation to bootstrap their development and alleviate the bottlenecks and constraints of capturing real-world data,” Huang explains. “Not every company has the time or the resources to rely on human demonstrations only. Simulation will continue to play a critical role in robotics.”
Robert adds that growing labor shortages and the need for uninterrupted operations are pushing companies beyond automation toward fully autonomous solutions.
“In today’s landscape, industry leaders increasingly view a cost-effective humanoid as not just an advantage, but a necessity,” says Robert. “Our vision is to build an autonomous future, and AEON is our flagship product in executing that strategy. We expect to introduce several variants in the years ahead, all in pursuit of that vision.”
Login to add comment
Other posts in this group
Rather than the Sunday scaries or toxic bosses, employees have unlocked a new workplace fear: office chair butt.
While not a new concern, the term has resurfaced on TikTok to describe ho
Music streaming service Deezer said Friday that it will start flagging albums with AI-generated songs, part of its fight against
In 2023, the flagship reveal at Apple’s WWDC keynote was unquestionably the debut of
What happens when venture capital and government pull back from science entrepreneurs at the same time? Many scientists think we’re about to find out, and are looking at how we can preserve our co
Research on misinformation and disinformation has become
