Humanoid robots won’t live up to their universal promise, argues Anton Duisterwinkel.
A marathon for humanoids. Movies of dancing groups of humanoids. Humanoids performing tasks at a BMW factory. Neura, a European humanoid startup, attracting more than a billion euros in investment. The humanoids are coming! Will they take over all physical tasks performed by humans, including in semiconductor manufacturing? And is that prospect a dream come true or the worst nightmare ever?
Humans are drawn to humanoids as bees are to flowers. Where normal robots evoke negative feelings (“tech that’s taking our jobs”), humanoids evoke sympathy through their shape and slightly clumsy behavior. Recently, the Humanoid Application Center was opened in Schiedam. More than 250 guests had beaten the long waiting list and were keen to see humanoids in action. Over ten companies have already signed contracts for testing humanoids for their use cases at the center.
We’re closing in on the peak of the Gartner Hype Cycle. Within a year or so, people will start to realize that current humanoids aren’t even close to delivering the promise of taking over all physical tasks. Consider paving a street with bricks: After lifting a brick five times, the poor thing overheats. Kneeling to pave a street is extremely difficult to program. And what happens when it starts to rain?
For such a task, a robot on wheels or caterpillars surely is cheaper, more stable and quicker. Also, the toughest job isn’t laying the bricks but placing the heavy curbstones. That’s clearly no job for a standard humanoid. The promise of one humanoid, or platform, for all use cases is simply ridiculous. On the other hand, the number of platforms – tailored for different applications – needs to be limited to reduce robot cost and speed up innovation.
As in any technology following the Gartner Hype Cycle, a certain number of applications of real humanoids will be realized in upcoming years. My best guess is that this will happen in tasks where human responses are expected or even needed. Hospitals and hotels, for example.
In many other cases, dedicated robots that use aspects of humanoid technology will be sufficient. Understanding which tasks should be robotized and how this is best done is no easy feat. It involves a good understanding of work processes and of psychological, social and organizational aspects. Researchers at Delft University of Technology and TNO, collaborating at the Fraim innovation center, are getting to grips with this ‘worker-centered robot development’ and will share their lessons learned with field labs all over the Netherlands. Professor Jessie Koen (TNO, University of Twente) has shown that even after introducing robots to a workplace, an adaptive process is needed to get the best out of workers and robots.
When we’re able to combine this humanoid technology with worker-centered robot development in domains that we really understand, like semicon, agriculture and maritime manufacturing, the Dutch robot industry can grow quickly. The use of hardware and software platforms will enable this quick growth.
However, we need to move quickly. Cheap Chinese robots, humanoid or not, will flood global markets in a few years. The window of opportunity is already starting to close. Robots are coming, surely, and some of them may take a humanoid form.
