From Automation to Autonomy: How Hexagon Robotics is Bringing Humanoid AI to the Factory Floor

Session: From Automation to Autonomy: Deploying Humanoid Systems in Real Industrial Environments.

Speaker: Darius Wilke, Vice President Commercial, Hexagon Robotics GmbH

Session Summary: Hexagon Robotics is deploying wheeled industrial humanoid robots, specifically AEON, to seamlessly integrate into existing human-centric factory layouts and solve critical manufacturing labor shortages. Instead of complex manual coding, these robots are trained through AI and imitation learning, allowing robust task models to be quickly learned from human demonstrations and instantly scaled across global fleets. Because industrial AI relies entirely on strictly guarded proprietary data rather than public information, manufacturers must prioritize early data collection to secure their future competitive advantage.

The modern manufacturing sector is undergoing a profound transformation. As factories transition toward highly adaptable production lines, the era of static, single-purpose robots is giving way to a new paradigm: the industrial humanoid. At the recent Hannover Messe trade fair, Darius Wilke, Vice President Commercial at Hexagon Robotics, outlined the company’s pragmatic approach to deploying humanoid robotic systems in real industrial environments. Distancing his company from viral internet videos of acrobatic robots, he emphasized a strict focus on tangible, value-driven work, noting, "our stage is the factory".

The Catalyst for Humanoid Automation

The shift toward humanoid robots is driven by several converging industrial challenges:

• Dynamic Production Needs: Traditional manufacturing relied on "high volume, low mix" production, where assembly lines ran unchanged for years. Today, industries are moving toward "low volume, high mix" models that require machinery to rapidly switch between different tasks and components.
  

• Human-Centric Workspaces: Most factory floors are ergonomically designed for human workers. Introducing traditional industrial automation often requires completely redesigning these physical spaces. Humanoid robots, however, can seamlessly integrate into existing layouts, using the same tools and workflows as their human counterparts.
  

• Labor Shortages: Companies are increasingly struggling to staff multiple shifts, particularly for mundane or physically demanding tasks. Humanoids present a viable solution to bridge this critical workforce gap, offering German companies a way to stay competitive against international competitors.

Meet AEON: The "Swiss Army Knife" of Robotics

Hexagon Robotics’ flagship humanoid, AEON, is engineered to be a versatile combination of an autonomous mobile platform and an industrial robotic arm. AEON stands 165 centimeters tall and weighs 60 kilograms. Notably, instead of walking on mechanical legs, AEON operates on wheels. This design choice allows the robot to navigate vast factory floors at speeds of up to 2.4 meters per second—significantly faster than the average human walking speed of 1.3 meters per second.

AEON is equipped with dual batteries that provide roughly four hours of continuous operation. The robot autonomously navigates to a charging station to swap depleted batteries for fresh ones, essentially taking its own "coffee break" to ensure uninterrupted labor. Its payload capacity of 8 to 15 kilograms is perfectly optimized for typical industrial tasks, as most parts handled by factory workers weigh under 5 kilograms to prevent ergonomic injuries and back pain.

The AI Engine: Imitation Learning

Programming a complex humanoid robot with manual code to coordinate arms, elbows, and hands would be exponentially time-consuming and impractical. Instead, Hexagon Robotics relies on an advanced Artificial Intelligence (AI) concept known as "imitation learning".

Human operators wearing Virtual Reality (VR) glasses physically perform a task—such as gripping and sorting parts—while the robot perfectly mimics their movements in real time. By observing these physical demonstrations across 10 to 50 "episodes," the underlying AI model mathematically generalizes the core objective. This repetition allows the AI to filter out human imperfections; if an operator's hand shakes during a demonstration, recording multiple episodes averages out the inconsistencies, resulting in highly stable, robust robotic motion. Consequently, a single hardware platform can be organically taught dozens or hundreds of distinct capabilities.

Fleet Scalability and the "Humanoid Gym"

A major advantage of AI-driven humanoids is their scalability across global enterprises. Robots are largely trained in digital simulation environments. Once an AI model is perfected, it is packaged as a "policy"—akin to a digital reflex—and can be instantly deployed across a massive fleet of robots.

For example, one client with a global network of 100 factories built a "humanoid gym" to specifically train models for their 20 most common operational applications. Once these policies achieved high robustness in the gym, they were deployed to a worldwide fleet of robots, entirely eliminating the need to train individual machines from scratch for those specific tasks.

The Critical Role of Proprietary Data

While consumer-facing AI systems like ChatGPT are trained on vast amounts of publicly available internet data, industrial AI relies strictly on proprietary data. Factory workflows involve tightly protected trade secrets, patented products, and highly competitive methodologies.

Because of this, industrial companies must build and train their own unique AI models. Premium manufacturers like BMW and Schaeffler, which actively partner with Hexagon Robotics, began experimenting with this technology early to build robust data sets. A company like BMW has exceptionally high expectations and would never share its highly robust training data with a direct competitor like Audi. Consequently, forward-thinking manufacturers are recognizing that early data collection is a crucial competitive advantage. The future of automation will not be manually programmed; it will be almost entirely data-driven.

The Path to Deployment

For companies looking to integrate humanoid robotics into their operations, Hexagon Robotics recommends a highly disciplined, structured approach:

• Targeted Focus: Avoid trying to automate the most technologically exciting or complex tasks immediately. Instead, target one or two highly repeatable, simple, and mundane human-centric stations.

  
  

• Use Case Discovery & Alignment: Hexagon Robotics collaborates with clients to review factory workflows and narrow down a broad list of potential applications into a highly viable shortlist of two to three achievable tasks.

  
  

• Lab Pilots: Customer manufacturing parts are shipped to Hexagon Robotics' engineering headquarters in Zurich, Switzerland, to rigorously test the feasibility of the proposed automation in a controlled environment.

  
  

• Factory Pilots: Following lab success, the robot is deployed directly onto the customer's factory floor, integrating into existing digital systems for continuous, semi-productive testing before moving into full, real-world labor.

While humanoid technology is profoundly disruptive, Darius Wilke cautions that it remains in its early stages and will take another 6 to 18 months for AI models to fully replicate complex human dexterity. However, by starting early and focusing on robust data collection, manufacturing companies can secure a lasting competitive edge in the incoming era of physical AI.