Where are soft robots?

What are some real world applications of soft robots? 

Soft robotics is very much a nascent field. Most of the activity is academic in nature, and only few companies have found product market fit. The company "Soft Robotics" is one of the better known soft robotics company. They have developed soft pneumatic grippers for the food industry where grip adaptability is key [2]. Another company, "Empire robotics" also leveraged soft robotics' adaptability to create a universal robotic gripper. Despite their gripper impressive capabilities [5], the company eventually failed to find market traction, leaving a very interesting real world application case study [1]. The company "Syntouch" on the other hand, seems to be focused on using soft sensors to deliver automation in the field of surface sensing and characterization [3]. 

Apart from automation, soft robotic technologies might also be a good approach for the design of soft exoskeletons (a.k.a. exosuits). Indeed, exosuits are by definition in close contact with humans, which makes them prime candidate for safer and softer technologies. Furthermore, demographic trends in developed countries point toward an aging population, indicating both possible tight labor markets and a high number of elderly people. As a consequence, work or mobility assist exosuits are expected to be relevant in the future. Another application of exosuits is rehabilitation, and numerous companies [4,7] have developed products aiming in that direction. A good overview of the exosuit and exoskeleton market can be found on the website exoskeletonreport.com [6]. 

If you subscribe to the idea that exoskeletons are a suitable application for soft robotics, then a striking particularity of  numerous of above mentioned exoskeleton designs is the incorporation of a large number of rigid elements equipped with soft attaches and pads for the user. This is in part due to soft actuation technologies being not mature and power dense enough to fulfill exoskeleton requirements. But this is also due to the requirement for some exoskeletons to transmit forces away from the user's skeleton, which can not be done using only soft technologies. Why do I point that out ? Soft and traditional robotics exist on a spectrum, from soft to rigid. I do not expect fully soft robots to meaningfully permeate our world in the short and medium term. I would rather expect hybrid designs based on hard and soft elements, similarly to how our own skeleton and soft tissues architecture. 

Should we be scared of soft robots?

Overall, current soft robots are research curiosities that are fairly weak. So, you shouldn't be afraid of them, yet. In theory, soft robots have lower weight and force than traditional robots, and in that sense, are less likely to injure someone than traditional robots. So, if you don't spend your life afraid of robots, then these new and soft robots should not keep you up at night. 

However, one should note that a soft robot is not necessarily safe [8]. Softness oftentimes being borne out of elasticity, future soft robots design will have to account for stored elastic energy. Indeed, a rubber band is soft and lightweight, it can still be painful when it snaps back from an highly extended state. 

 Ultimately, I think it's safe to assume that any robot that can do significant work can be somewhat dangerous. Bugs and edges cases will always exist, and integrating robots within our day to day life will engender risks. As demonstrated with automated driving cars, the main question then is not "Is this robot safe?" but rather "Is this robot safe enough?". When applicable, a robot relative safeness may also be measured against it's human counterpart. For example, one could rationally argue that self driven cars should only be demonstratively safer than human drivers to be "safe enough". Overall, we need to accept that that risk zero does not exist. Lawmakers, citizens, and governments will have i) find ways to accurately measure risks and benefits, and ii) to choose when the benefits outweigh the risks. Using such a framework should help decide when and where to further automate our environment.