From Robot Arms to Humanoid Robots

ICT Group has been applying its expertise as an embedded software integrator to factory automation for many years. That same expertise is proving equally valuable in robotics applications. As a result, the High Tech division recently expanded its focus to the robotics market, a move underlined by its membership of NL Robotics, the Dutch robotics industry association. By doing so, ICT Group helps organizations translate technological progress into tangible business value. We spoke with Koen van Wijk, Technology Lead within the High Tech division.

Humanoid Robots

Van Wijk’s fascination with robotics goes back to his Electrical Engineering studies in Eindhoven, where he participated in robot competitions. His children inherited the same enthusiasm and began experimenting with LEGO robots several years ago. “They took part in the FIRST LEGO League, a competition where teams build and program LEGO robots to perform a variety of tasks,” Van Wijk explains. “When my children became interested in building humanoid robots a few years ago, I started diving deeper into the technology myself and visiting trade fairs to see what was already possible in practice.”

While robotic arms excel at repetitive work, humanoid robots are capable of handling a much broader range of tasks. Media demonstrations already show robots folding T-shirts, performing backflips and dancing. For many organizations, however, the challenge remains determining exactly how these capabilities can create business value.

Physical AI

Van Wijk considers Physical AI one of the most significant technological developments in robotics in recent years. He believes generative AI will have a similar impact on robotics as it is currently having on software development. “Last year, developers were still writing most of their code themselves with some AI assistance. Today, AI agents can autonomously create complete software solutions. In the future, we will simply attach a camera to a robotic arm and, with minimal programming, an AI model will translate camera images and sensor data directly into actions that the robot can execute independently.” Physical AI is currently a major research area for companies such as NVIDIA, which are training AI models specifically designed for interaction with the physical world.

Generative AI is also becoming increasingly important as the interface between humans and machines. If an engineer needs two days to teach a robot a new task, adoption remains difficult. According to Van Wijk, it will soon be possible to instruct a robot through natural language or by demonstrating a task. A user could simply tell a robot to place a specific screw in a different location and the system would understand and execute the request. “Just as AI agents such as OpenClaw have opened up entirely new possibilities in software development, we will see similar AI agents emerge at the planning and control level of robotics.”

“As more robots are produced, costs will continue to decrease, making them viable for a much wider range of applications.”

Accessible Robotics

Another important trend Van Wijk identifies is the increasing accessibility of robotics technology. “Today, the business case for many applications is still challenging because robots remain relatively expensive. That cost is not only related to the purchase price, but also to the time required to reconfigure them for different tasks. However, as more robots are produced, costs will continue to decrease, making them viable for a much wider range of applications. Companies will no longer need production volumes of 100,000 units to justify automation; even smaller volumes will become economically attractive.”

“With cheaper hardware and an open-source software stack, it is only a matter of time before every company can start working with robots.”

On the software side, Van Wijk also sees rapid acceleration. Hugging Face, the company behind the well-known open-source AI community, is developing LeRobot, a complete software stack for robotics. “They provide AI models, datasets and tools that allow robots to learn from videos by imitating observed movements.” Combined with increasingly affordable hardware, these developments are lowering the barriers to robotics adoption.

Van Wijk expects the first major breakthroughs to occur in logistics and assembly environments. “Sorting parcels and packing and unpacking goods is becoming increasingly accessible. Mobile robots that move autonomously through factories have also reached a relatively mature stage. After all, factories are controlled environments.”

“Through our experience in factory automation, we know just how critical safety is.”

Safety

Despite the rapid progress, significant technological challenges remain. “Safety is far from a solved problem,” Van Wijk warns. “Humanoid robots can weigh up to 35 kilograms. Imagine such a robot falling onto an employee during a power outage. Through our experience in factory automation, we know just how critical safety is.”

“The business case for any application ultimately depends on the ability of AI models to apply the knowledge and skills acquired during training to new situations.”

Another key question is how broadly Physical AI models can generalize. “If you have to demonstrate how to fold a T-shirt a thousand times and the robot can only fold T-shirts afterwards, the value is limited. But if the same model can also perform many other tasks, then the training effort becomes worthwhile. The business case depends on the extent to which AI models can transfer what they have learned to situations they have never encountered before.”

“The question is therefore not whether robotics will break through, but how rapidly organizations will embrace the opportunity.”

More than just a robot

A robot never operates in isolation; it is always part of a broader system, Van Wijk emphasizes. “As a system integrator, we add value by helping customers identify the robots that best fit their needs and integrating them into their existing systems. That involves complex disciplines such as embedded software, real-time processing, safety and AI. This combination is exactly where our strength lies. We possess deep technical expertise in all these domains, combined with broad experience across multiple industries. We are already seeing concrete requests for robotics solutions in high-tech environments and expect to move quickly to the next stage. The question is therefore not whether robotics will break through, but how rapidly organizations will embrace the opportunity.”