More evidence that humanoid robots’ ‘natural’ home is in the automotive sector arrived this week, at NVIDIA’s GTC Berlin AI conference, which wraps today.

At the event, Munich, Germany, based Agile Robotics announced a four-way partnership with upscale carmaker Audi, France’s Dassault Systèmes (the industrial software member of the Dassault Group), and $5 trillion US hardware behemoth NVIDIA. The aim is to explore automating key logistics processes within Audi’s manufacturing plants.

At the centre of the partnership is an unnamed mobile humanoid, which Agile is developing for the intralogistics stage of a vehicle’s assembly. The robot is being trained at Audi’s Production Lab to pick and sequence different parts from containers, ensuring that the right ones reach the right workstations in the correct order.

At present, these slow, complex, but routine and repetitive processes are carried out by human workers, who have the expertise to pick and sequence the parts from containers of thousands of unsorted items.

However, this critical function has long been a bottleneck in the assembly of premium vehicles, with more than one million picks daily across Audi’s production lines. The thinking is that by automating them 24×7, using specially developed, dextrous humanoids, the bottleneck could finally be removed, helping Audi to produce its cars more quickly and efficiently.

The partners each have clear roles in the relationship: Audi provides the training, human expertise, and deployment, Agile Robotics the humanoid platform, NVIDIA its Omniverse simulation and collaboration libraries for Physical AI, and Dassault Systèmes a virtual twin of the logistics environment itself, built using its DELMIA system.

Along with teleoperation, simulation of this kind is the critical technology in helping humanoids and other robots cross the so-called Robot Data Gap quickly.

Put simply, the data needed for robots to understand the physical world is not available on the internet. As a result, it can only be sourced from the real world directly – by teleoperation, for example – or by training in virtual environments. Via simulation, complex actions can be trialled and optimised more quickly, and robots allowed to fail safely until they are able to carry out each task reliably and repeatedly.

Not human replacement 

The new Agile robots themselves are human-sized, with a sensor-filled head, two arms, and grippers, but they each have a mobile base, rather than legs, which, effectively makes each of them into a humanoid and AMR (Autonomous Mobile Robot) hybrid. AMRs are commonplace in the manufacturing and logistics sectors for transporting containers from one location to another in unstructured spaces.

Unlike some other applications of humanoids in the automotive sector, the aim is not large-scale human replacement, claims Audi, but a collaborative partnership between expert engineers and robots, with the latter performing the repetitive, low-value-adding parts of the job.

In this way, the deployment is analogous to lab automation in life sciences, where highly qualified scientists often spend inordinate amounts of time performing routine and exhausting tasks which, if automated, would free them up to do more high-value research.

Making the announcement, Agile Robots’ founder and CEO Dr Zhaopeng Chen said:

Agile Robots is driving industries forward with Physical AI. Working with Audi we show how our mobile humanoids can perform high-value, critical tasks that were previously impossible to automate. That is what Physical AI is all about – real progress for industrial productivity.

Gerd Walker, Board Member for Production and Logistics at Audi AG, part of the Volkswagen group, added:

The future of manufacturing will be defined by how successfully we combine industrial expertise with Artificial Intelligence. For decades, picking within the sequencing process has remained one of the toughest automation challenges in automotive logistics.

Also at NVIDIA GTC Berlin, Agile Robotics – not to be confused with US vendor Agility Robotics, maker of the Digit humanoid – announced a new variant of its Agile ONE robot.

The compact, 1.4m Agile ONE SN, built on NVIDIA’s Isaac GR00T platform, is intended purely as a research and development robot, allowing developers to build, train and validate humanoid skills in lab conditions before deploying them in the real world.

Transports of robo-delight

As previously reported on diginomica, the automotive sector is proving to be both the ideal test case and the perfect deployment environment for humanoid robots. Indeed, they have a much more promising future in manufacturing than they do in domestic applications, where the Robot Data Gap will be much harder to fill safely and reliably.

In September, Japanese car giant the Toyota Motor Group stunned the industry by announcing plans to deploy 400,000 humanoid robots in its own factories, working alongside expert human staff – at least initially.

Toyota aims to begin rolling out fleets of its in-house developed ELEY robots in 2028, as part of a factory modernisation program. However, ELEY is explicitly designed to learn and replicate human tasks in situ, and then to share that data with the fleet. In this way, the company appears to be building an automated infrastructure for human replacement, rather than expert task augmentation.

Also in September, rival Mitsubishi Motors announced that it is converting a disused engine factory into a humanoid production facility, with the reported aim of producing 1,000 robots a month next year.

Mitsubishi has signed an agreement with Highlanders Inc, a robotics startup from the University of Tokyo, to jointly develop the humanoids, which will initially be deployed in the carmaker’s own factories, before being offered to external customers.

Meanwhile, the big news at the Consumer Electronics Show (CES) back in March was all about humanoids too. At the event, Boston Dynamics launched the all-new, electric product version of its Atlas robot, familiar to many from videos of earlier hydraulic prototypes running, jumping, and turning somersaults.

But the real story for enterprise buyers was Atlas’ immediate future in the Hyundai Motor Group, which has been Boston Dynamics’ majority owner since 2020, and its controlling shareholder since 2021.

Atlas is being trained to perform routine tasks at Hyundai’s Robotics Metaplant Application Center (RMAC), and at AI partner Google DeepMind’s research facility. The medium-term aim is a rollout of the robots at the Hyundai Motor Group Metaplant America in 2028.

But in the longer term, Boston Dynamics said it plans to deploy “tens of thousands” of the robots in Hyundai’s factories, and to open production lines capable of building 30,000 robots a year.

Also in the Spring, Microsoft announced a strategic partnership with Stockholm-based Hexagon Robotics to accelerate humanoids’ advance in the manufacturing sector.

At the center of the collaboration is Hexagon’s AEON humanoid, which is designed to operate safely and autonomously in complex industrial settings. Like several other humanoids, AEON runs on NVIDIA’s Jetson Orin processor and was largely trained via simulation on NVIDIA’s Isaac platform.

In March, the BMW Group began trialling Hexagon’s robots at its factory in Leipzig, Germany, working on battery and component assembly lines. In 2024 and 2025, the company also trialled Figure’s 02 humanoid at its US plant in Spartanburg, South Carolina, where the robot positioned sheet metal parts for welding.

Other robot makers are finding their natural partner in the automotive industry too. In February, Salem, Oregon-based Agility Robotics signed a commercial agreement with Toyota Manufacturing Canada to supply its Digit humanoid as a support robot in manufacturing, supply chain, and logistics operations.

Making the announcement, Agility Robotics’ CEO Peggy Johnson said:

With our next generation of Digit [Digit5, launched during the summer], we will be the first company to deliver the first cooperatively safe humanoid robot to work alongside people, allowing companies like Toyota to scale their use of humanoids well beyond what is possible today.

Amazon is among the many customers that are trialling Digit in logistics support settings, where robots such as Sanctuary AI’s Phoenix and Apptronik’s Apollo humanoids are also destined to work in the medium term.

My take

One thing is clear: if humanoid robots have anything like the future that is envisaged for them by technologists – and by their investors – then automotive will be the proving ground.