Cracking the automated wire harness assembly challenge
Developments in machine vision, force control and a modular robot cells could finally close the automation gap for wire harness assembly
Automotive manufacturing bodyshops, paintshops and battery production can and do reach high levels of automation, but wire harness production remains fundamentally different: flexible wires, multiple branches, small connectors and very high variant complexity still make this vital component a genuinely difficult product for machines to handle.
Cellios, a spin-out from the Fraunhofer Institute for Manufacturing Engineering and Automation (IPA) in Stuttgart, together with project partner TE Connectivity and machine vision specialist MVTec Software, has now developed what is claimed to be the world's first fully automated wire harness assembly system – a development with implications not just for automation engineers, but also for automotive supply chains.
The wire harness automation challenge
The scale of the challenge Cellios set out to solve is significant. Wire harness manufacturing has traditionally been what the industry calls a "migratory industry" – production that continually shifts to countries offering the lowest labour costs, precisely because the process has remained so resistant to automation. That resistance comes down to the physical nature of the component itself. Commenting on a similar, if smaller scale project with the University of Tennessee Nissan Industrial Strategy Engineer, Aaron Hall, described the same fundamental problem: "A wire is not a rigid structure like you would see in body assembly or general assembly. It's unpredictable and almost uncontrollable unless you have a grasp on the entire thing." Unlike a rigid part on a conveyor, a cable's exact position changes with every handling operation – a fundamentally different problem for a robot than picking and placing a fixed object.
Previous solutions often covered only partial steps of cable processing, meaning that pre-assembled cables had to be further processed manually. The goal of the project with TE Connectivity is to fully automate the entire process, right up to electrical testing
The wider industry is grappling with this challenge from multiple directions. Hyundai Mobis has been running its own pilot project at its Ulsan Electrification plant in South Korea, focused specifically on automating the attachment of wire harnesses to EV battery modules – a related but distinct step in the value chain. Elsewhere, the Next2OEM research project involving Audi has demonstrated that when a harness and its production process are designed together from the outset, automation levels can rise from around 10% to roughly 90% within a manufacturing demonstrator – though that project focused on redesigning the product and process jointly, rather than automating an unmodified, traditionally designed harness end-to-end as Cellios has set out to do.
Cellios' work with TE Connectivity goes further still, aiming to automate the entire wire harness assembly process, rather than a single connection point or pre-assembly stage. "Previous solutions often covered only partial steps of cable processing, meaning that pre-assembled cables had to be further processed manually. The goal of the project with TE Connectivity was therefore to fully automate the entire process, right up to electrical testing," explains Dr. Frank Nägele, CTO at Cellios.
The precision problem: inserting a crimp into a connector
At the centre of the challenge is a component just a few millimetres wide: the crimp, the terminated end of a cable that must be inserted into a corresponding connector cavity. Because the cable itself has no rigid shape, the exact position of the crimp within the robot's gripper shifts with every handling operation – meaning the robot cannot simply rely on a pre-programmed, fixed trajectory to complete the insertion.
To reliably guide the crimp into the connector cavity, the system requires an accuracy of at least one-tenth of a millimetre. That tolerance sits at the outer edge of what industrial robotics can typically achieve without additional sensing and correction. "In addition to the technical issues, there are also complex process-related questions. After all, the objective is to automate a process that has been entirely manual. Components such as connectors, cable-processing machines, and so on are designed for use by humans, not robots. On top of that, there is an enormous variety of wire harness variants that we want to accommodate through the flexible design of our assembly cells," says Dr. Nägele.
A modular system, broken into discrete process steps
Cellios' solution is built around modularity, mapping the different stages of wire harness assembly to individual robot cell modules. These process steps include connector singulation, cable preparation, crimping, cable routing, producing splice connections using an ultrasonic welding device, taping, and finally end-of-line electrical testing.
Of all these stages, routing and contact insertion have proven the hardest to automate. Historically, this task has been performed entirely by hand, directly on the routing board, and has demanded intense concentration, visual acuity and manual dexterity from operators. In Cellios' newly developed robot cell, the robot instead moves the gripped component to a 2D camera, which determines the crimp's position and orientation and sends that data back to the robot as a real-time correction. A second camera then measures the target position on the connector side. Using these two data points, the robot performs precise corrective movements to reach its target and reliably insert the crimp.
Precision alone is not sufficient, however – the insertion also requires a degree of physical sensitivity that vision systems cannot provide on their own. This is achieved through force control and force monitoring: a force-torque sensor mounted on the robot measures the forces generated during insertion and regulates them in real time, ensuring the crimp is inserted correctly without damaging either the crimp or the connector.
Without precise measurement, we would not be able to insert components with an accuracy of 0.1 mm, which is what we need to align the connector with the crimp
Giving the robot "sight"
The system's ability to see and correct for these variations depends entirely on machine vision. "Without precise 'vision', the robot would literally be blind and unable to make the necessary corrections reliably and independently," Dr. Nägele explains. Standard sensors alone were not capable of determining the exact position (X-Y) and orientation (Z angle) of the crimp with sufficient reliability, which is why Cellios turned to a dedicated machine vision system to identify deviations introduced by upstream process steps – such as the crimping of the cable itself – as well as the unpredictability of the cable's physical form.
"Without precise measurement, we would not be able to insert components with an accuracy of 0.1 mm, which is what we need to align the connector with the crimp," Dr. Nägele adds.
The robot cell's hardware reflects the complexity of the task: alongside the robot itself, the system incorporates robotic taping applicators, force-torque sensors, purpose-built connector-gripping devices, and a conveyor system, with two 2D cameras handling image processing. On the software side, the system combines Cellios' own proprietary software with MVTec Merlic, a machine vision platform developed by Munich-based MVTec Software GmbH.
Why Merlic – and what a low-code platform enabled
Cellios selected Merlic specifically for its low-code approach to application development. Rather than requiring specialist programming or image-processing expertise, Merlic's configuration interface allows the required vision tools to be assembled through drag-and-drop selection.
Within Cellios' image-processing workflow, Merlic's matching technologies establish the exact position and orientation of each crimp, while built-in MQTT and REST interfaces provide integration into the wider robot cell system – extending the software's role beyond core machine vision functions into broader system communication. The platform's accessibility proved a practical advantage during development: two Cellios employees implemented the vision application using only instructional videos and sample applications, without requiring additional formal training.
"Our collaboration with Cellios is a useful demonstration of how machine vision is pushing the boundaries of what is possible in robotics. When robots learn to see with the same precision and flexibility as humans, even highly complex processes can be automated," says Ulf Schulmeyer, Merlic Product Manager at MVTec.
Bringing quality upstream
Cellios' approach also reflects a broader shift happening across harness production: moving quality control further upstream in the process, rather than relying solely on end-of-line inspection. A poor crimp or incorrectly inserted terminal can become expensive if discovered only after final installation. By capturing process data directly during insertion, monitoring each step and confirming component positioning via camera, Cellios' system embeds inspection into the assembly process itself – digitally monitoring every insertion operation and enabling complete traceability of the finished harness, a capability that is difficult to replicate consistently under a fully manual process.
From prototype to market — and what comes next
The prototype system was presented in November 2025, with an official market launch planned for early 2027. The implications of the technology extend beyond a straightforward productivity gain on the factory floor. By automating a process, the system has the potential to make wire harness production economically viable again in high-wage countries – reversing, at least in part, the migratory pattern that has defined this segment of the supply chain for decades.
Based on early customer feedback and ongoing discussions, Cellios is already planning to extend the automation approach to additional wire harness variants, as well as into adjacent applications such as control cabinet wiring. The company also expects future iterations of its machine vision systems to support self-healing processes and more advanced quality inspection capabilities.
Systems and process development are continuing to improve the viability of this ambitious automated approach. At this stage of development, the processes to ensure the accurate installation of the connectors means it is slower than a manual operation but Dr Nägele notes that “the system that we demonstrated at Productronica 2025 (https://www.youtube.com/watch?v=I-8ijqbiPTk) is still relatively slow, however, we are working on our first production system which will start producing harnesses by the middle of 2027. This will be 4 times faster (through parallelising processes and optimisation) and will match the speed of the manual operations.”
He also highlights plans to reduce cycle time in the bottleneck of the system (routing cables), which will make the second production system (2028) even faster.
Scaling for production and scope for the future
Another important area for development is scaling this technology from project to volume production. Dr Nägele explains that this is a key part of Cellios' planning. “Each system will produce harnesses for large scale automotive volumes (100.000+ harnesses/year).
“Scaling up our company to be able to produce multiple robot systems per year will require automating much of the engineering, manufacturing, integration and test efforts to design and build the systems, but we are already working on this.”
Cellios' achievement arrives at a moment when the wider industry is converging on the view that better robotics alone will not solve wire harness automation – connector geometry, wire routing, zonal E/E architectures and digital product data all need to evolve in parallel. Companies such as BMW have already demonstrated that zonal wiring architectures can cut harness length by hundreds of metres and reduce harness weight by around 30%, easing the physical burden that any automation system, robotic or otherwise, ultimately has to contend with. Whether or not Cellios' fully integrated approach becomes the industry template, its success illustrates that the wire harness assembly – long considered manufacturing's most stubbornly manual product – may finally become a fully automated process.
