Artificial intelligence: time to 'invest in soft robotics'
by Rolf Pfeifer
The increasingly popular field of soft robotics is set to have an enormous impact on the service and manufacturing industries; and there is no need to be concerned that automation will put people out of work – the reverse is true as jobs that would otherwise be outsourced to China can be protected
Is soft robotics a fad, or the future? If you browse robotics-related internet sites, you will find a plethora of workshops, conferences, summer schools, and publications on the topic of soft robotics. What is it, why is it becoming such a popular topic, and what are the implications for the future of technology and society? The concept of 'soft' in robotics applies to multiple levels: literally soft to the touch – skin, tissue on body surface, muscles, tendons – soft as in natural movements in contrast to manufacturing robots or many walking robots, and soft as in safe interaction with humans.
Looking at current state-of-the-art robots, especially in manufacturing, they are usually made of hard materials, metal and plastics, and equipped with very fast, precise, motors. Manufacturing environments are highly predictable. Normally, nothing unexpected happens and the robots can be programmed down to the last detail by applying methods from control engineering. By contrast, biological systems, such as humans or animals – though they have hard bones to provide structural support – are mostly made of soft materials, have soft skin, and soft tissues on the body surface, behind which you find soft, elastic muscles and tendons. They have to function in real-world environments which are characterised by high degrees of uncertainty and rapid changes, such as a shopping centre, a school, a home, an entertainment park, or a football field.
Soft materials, given the proper shape, or morphology, have highly interesting characteristics from a robotics perspective, because they can incorporate part of the control – a process which is known as 'morphological computation'. An elastic muscle can cope with unevenness of the ground and impact in walking, the stiffness of the tailfin in fish optimally distributes the forces over the fin as it oscillates back and forth, and the soft tissue on the fingertips and in the hand passively adapts to the shape of a hard object, such as when drinking from a glass.
While materials provide great advantages in terms of 'easy' and 'cheap' additional functionality, something dramatic happens here: the clear separation between control and the to-be-controlled breaks down – at least partially. Moreover, soft materials tend to be much more difficult to model than hard ones, so that traditional control methods no longer apply. In other words, entirely novel design concepts will be required; principles that now must include morphological and material considerations, which, of course, leads to an explosion of the design space – which, in addition to the computational aspects now includes morphology and materials.
But the hope is that the functionality required for real-world environments where reactivity and adaptivity to rapid unpredictable changes is crucial, can be achieved more effectively by exploiting the entire design space. It turns out that if the constraints that can be gained from the embodied nature of such systems are properly exploited, on the one hand the dimensionality of the design space can be reduced by orders of magnitude, and on the other, additional functionality can easily be achieved – again, an instance of morphological computation.
We expect soft robotics to have an enormous impact on the service robotics industry because we will share our living space with these machines, and we will closely interact and cooperate with them. This implies that they have to be safe and the interactions smooth and soft; safety has to be guaranteed, even if there is a power outage and the active control no longer functions. Speculating a bit further, we can also expect an enormous impact on the manufacturing industry because tasks that have defied 'cheap automation', for example automobile production, can potentially be robotized by employing principles from soft robotics. New and innovative industries will become amenable to robotic technology.
A good example is the food processing area where varying, unanticipated shapes need to be dealt with, or generally manufacturing jobs, where close human-machine cooperation is required. Of course, soft robotics will not be the only technology, but combined with others, such as parallel robotics, just to mention one line of approach, we can expect a real boost. In fact, some people from the pertinent industries, Heartland Robotics and Adept, two leading American robotics companies, even predict a new industrial revolution.
Let us look at China for a moment. Foxconn, a Taiwanese technology giant, is producing – among many other things – iPhones and iPads in China, where they maintain a workforce of roughly 1.2 million people, half of whom are located in Shenzhen, one of the Chinese economic strongholds and one of the fastest growing cities in the world. With the increase in worker salaries – more than 20 per cent in 2011 – China is partly losing its competitive advantage as a cheap labour country. Moreover, for European or American companies, it may no longer be the best strategy to simply outsource manufacturing tasks that seem hard and expensive to automate, to China. Because with soft robotics – again combined with other technologies – the degree of automation can be dramatically increased, we would speculate that certain tasks can, in the future, be re-insourced to Europe or the United States, thereby raising the companies' competitiveness, and in turn maintaining people's jobs.
For a number of years, there have been negative media reports about Foxconn because of allegedly poor working conditions. As a reaction, in addition to a substantial increase in salaries, they plan to install one million robots over the next three years – without any job cuts whatsoever, moving people from dull or unhealthy jobs to more interesting ones, according to Foxconn founder and chairman Terry Gou. Whether or not these robots will incorporate soft technologies is, at this point, unknown – but they would definitely benefit from doing so. By the way, an interesting detail: Foxconn will open additional plants in Brazil, expected to create 100,000 new jobs.
It is time for western companies to wake up and invest heavily in robotics technology; otherwise, they will lose out against competitors like Foxconn. And forget about jobs being lost through automation – the exact opposite is the case, as documented by many economic and technological studies. As John Dulchinos, the chief executive of Adept, the largest US-based manufacturer of industrial robots, argued in a recent interview, the US has lost several million jobs in manufacturing to China because they did not automate their production lines but outsourced them to cheap labor countries.
The European Commission is doing an outstanding job at supporting basic research and development in the area of robotics and related fields with the 7th framework programme, and in the future with Horizon 2020. Examples of soft robotics basic projects include OCTOPUS, with the goal to build a robot octopus with its enormous dexterity and manipulation abilities, or ECCERobot, a fully tendon-driven, soft humanoid robot. Another prominent example is ECHORD, the European Clearing House for Open Robotics Development, which has the aim of bringing industry and academia – basic research – together.
Hopefully, the Future and Emerging Technologies flagship pilot 'Robot Companion for Citizens – for a sustainable welfare', a projected €1bn project, will be successful. Core pillars in there will be based on soft robotics, material technologies, and morphological computation. From this project, we cannot only expect scientific breakthroughs, but high-impact, disruptive spin-offs that will benefit the society at large. In conclusion: soft robotics – fad, yes; future, yes.
Professor Rolf Pfeifer is deputy director of the Swiss National Centre of Competence in Research Robotics and director of the artificial intelligence laboratory at the University of Zurich