Soft robotics is finally leaving the research lab and entering the real world. One of the companies leading the way is a Boston-based startup that is commercializing the innovations of the Whitesides Research Group at Harvard. I’m talking today with Soft Robotics CEO Carl Vause. Full transcript below.
Andra: I’m talking today with Soft Robotics CEO, Carl Vause. Can you tell us about your work?
Carl: Our expertise is in the soft actuator, where we are able to make robotic actuators of completely elastomeric material – rubber, silicon rubbers, polyurethanes. It’s a really different aspect of making what you’re normally used to … rigid manipulators, or parallel bar linkages, or linear actuators … but we’re really doing it without any rigid components.
Andra: And what research has this come out of?
Carl: The company was actually spun out of Harvard.
In 2007 there was a DARPA project called ChemBots and Harvard was an awardee on that project. Professor George Whitesides in the chemistry department actually took up the challenge… [which was to] build an octopus. And instead of trying to build a mechatronic robot that we’re all familiar with, they developed the soft actuator … trying to replicate the tentacle. Out of that came the intellectual property that is called ‘pneumatic networks’ – and using pneumatic networks you can build all kinds of different actuators to make fingers, hands, joints and the like.
Harvard worked on this for about five to six years, and then there was a decision that the technology had matured to the point where it was no longer in the best interest to continue research. [Instead they began] commercializing the technology and using it to solve some well-understood problems, and so the company was put together in the late spring/early summer of 2013.
Andra: What are some of the problems and specific applications that you are looking at for Soft Robotics?
Carl: One of the things that’s really interesting about the technology is that it’s low cost, it’s conformal, and you can design the hands in an open loop system to apply just the right amount of force.
The example I use is picking a tomato off of a vine. This would be a very, very difficult problem for a traditional robotic manipulator to do with closed-loop feedback. We can design a soft hand with elastomeric fingers that will only operate at a force that is below the crush force of a tomato. And so you just simply visualize a tomato, you activate the gripper, and it will grab the tomato and pull it.
This really opens up what I like to think of as the future of robotics into collaborative robotics: having robotics being able to work side by side with people and factories, handing instruments and tools to workers on assembly lines … or automating areas that we haven’t been able to automate until now, such as produce processing and food processing. Right now it’s very hard to handle fruits and vegetables with robots, but we really feel that soft robotics will unlock this capability.
Andra: You’ve mentioned collaborative robotics – the ability to safely work alongside people – as one of the significant benefits. And you also mention the opening up new areas, for example handling food and agricultural products. Do you think this will also apply to some of the other areas that have not been so easy for robotic manipulation, like fabrics, textiles, other deformable materials?
Carl: Absolutely I do.
One of the big challenges in robotics is how do you handle these soft things. If you’re shipping a shirt from the shirt manufacturer to an outlet like The Gap, they come folded in a plastic wrap. Picking up a folded shirt in a plastic wrap is very difficult for a robotic manipulator, but we’ve actually shown [we can] handle these delicate highly unstructured objects, and we can do it with a lower complexity than is typically done today. And that same grasper we demonstrated handling a folded shirt can also pick up a bag of rice, or things as varied as children’s toys.
Andra: I find soft robotics a very interesting area because of these wide applications, but naturally there are some limitations as well. I assume there are tradeoffs in accuracy or perhaps in strength. Where do you feel the boundaries of the applications are?
Carl: The biggest feedback we get right now is lifespan.
Traditionally in the research environment, soft robotic devices have not had the best lifespan – they’ve had very low cycles – and if you are talking about automating things such as textile handling, produce handling or collaborative robotics, you need to have cyclic lifetimes that are similar to traditional hard graspers. This is a problem that we’ve really focused on solving, and right now in our lab we have one of our soft manipulators running on a lifetime test. As of this morning we are approaching three million cycles, so we feel like we’ve really bridged that lifecycle/lifetime problem that some soft robotics technologies have had.
The other challenge we have had and are working to address is speed of grasp. Sometimes in soft robotics you’re moving a fluid – whether it’s pneumatic or hydraulic – and it’s obviously at a slower rate than with an electrical motor. But we’ve also demonstrated grasp times in the hundred millisecond range. We feel that the current limitations are well known, but that they can be solved and that solving them is deterministic.
Andra: Are there some advantages as well, in terms of the cost and power-to-weight ratios?
Right now we have a small gripper – a light-payload gripper – that can handle about 600 grams reliably. But it weighs only 100 grams, so when you put it on the end of a 1kg arm, you’ve got quite a bit more payload than a traditional device where you’re looking at maybe half of your end-of-arm payload taken up by the weight of the end-effector. We’re able to create very lightweight end-effectors, which make can collaborative arms much more productive in terms of payload capacity.
The other advantage that’s really been a powerful part of soft robotics is the cost. All of our cost goes into the engineering of the devices, but once we can get manufacturing a device, it’s really composed of proprietary design of the actuators, and then a light material such as polycarbonate and silicon elastomers. We’re not talking about an end-effector that’s got a lot of sensors and stepper motor and linkages in it. We believe that the standard cost for an adaptable end-effector can come down in price by an order of magnitude very easily, which is really important to opening up this new market of collaborative robotics.
Andra: You’re having to create a startup that’s going into unchartered waters, and I’m hoping you can tell me a little bit about the process. Soft Robotics has been in existence as a company for just over one year?
Carl: Since July of 2013.
Andra: How many people are there on your team?
Carl: Right now there are five of us. Two fulltime engineers, myself, and we use two consultants to keep the books and take care of our regulatory responsibilities.
Andra: Your background is not in robotics research so much as in business, and you’ve been doing medical devices before this. How did you come to join Soft Robotics?
Carl: My undergraduate degree was in electrical engineering and I worked on doing stepper motor control with 8-bit micro controllers. We were running assembly language at the time, and it was very miserable. I had forgotten all about that and was spending a lot of time with medical devices. I was introduced to [soft robotics] technology, and we were looking at various medical applications for it because of its soft conformal nature, obviously. The thing that makes soft robotics fantastic for handling tomatoes also makes it fantastic for handling human organs.
But when we got started, I remembered all the difficulties I had with control systems and feedback loops, and realized that we had a soft open-loop manipulator technology and that we should really spend time with the robotics community. And once we sat down with the robotics community, [their] feedback was, “This is the kind of thing we’ve had been waiting for! We’ve been trying to make the human hand, and that’s a very costly endeavor that takes a long time and a lot of effort … and you’ve created this highly dexterous, conformal, lightweight device in a short period of time!” With that initial feedback we decided to put all of our major efforts into focusing on the robotics opportunity.
Andra: Has this been supported through commercialization grants through the university, through the labs? What’s the process that has taken you this far, and what’s coming next?
Carl: It’s very difficult to run a hardware startup these days. We’re in the days of mobile apps and wonderful online portals that can access billions of people, but hardware startups require people and engineers and time.
We were lucky in that we did get some initial funding through the government, through two DARPA grants. DARPA had seen thetechnology in the university lab and asked us to push it forward. We’ve also been very lucky with angel investors. The company has been funded partially through government grants and the majority through angel investors who have seen the technology, done their fact checking with the robotics community, and who believe that this is something that will change robotics and will be very important going forward.
Andra: What I see is that angels are playing quite an important role in getting robotics startups closer to a large seed round or series A round, and it is a slower process than it is for many other startups. Certainly mobile and apps tend to give the impression of being cheap and easy to get off the ground, and it’s not totally the case. Whereas with hardware, it’s rarely the case that it’s cheap and easy.
Carl: It’s very interesting because in many of the discussions I have with investors, I try to help them understand that we need to buy materials, we need to fabricate things, we need to have access to a machine shop, and these are not things they’re accustomed to hearing from people working on software.
Andra: What do you think is coming up, that you’re facing?
Carl: The biggest obstacle in this is when you’re taking the technology out of university. The universities do very cutting edge early-stage research, and they’re not concerned about things like cost and manufacturability and scalability of a business process. That is what we spent the first nine months working on: could we get a good design space where we could build soft robotics in a 3D CAD system. That required a lot of work by our engineers. Second, we had to find vendors who could help us actually manufacture the parts in quantity, so that when we did start engaging companies we could sit down and say, “Look we’ve got a design space, we understand the design variables to solve these problems, and we have a manufacturing house that can make these in lots of 100 and lots of 1000 if desired.” That was a good nine months at the beginning of the company.
Now it’s really about understanding what the most important problem to solve is. We can do a lot of things with the soft gripper, and we’ve had suggestions for manipulating chickens and poultry processing, which is very interesting, but I know very little about poultry processing.
So we’ve been engaging the integrators, the end-users, and the robotics companies to really understand what problems they would like to see solved.
To circle back, the dominant themes really are realizing that dream of collaborative robotics, and automating industries (or parts of industries) that haven’t been conducive to robotics yet. But the feedback is, with soft robotics we could really bring automation to those new markets.
Andra: Are you going to tackle this one market at a time? For example, prove one out, and the add another? Or prove one out and then add ten or a hundred others?
Carl: I would say there’s maybe a third way.
What I’d like to do is to prove one out, focus on one market and really show that we can make a difference and then use the technology to go into new markets. And as early stage startup, focus really is important to our success.
As we are proving out these ‘one’ markets, one of the things we’ve been focused on is the unstructured warehousing environment: the ability to pack something that comes in a small electronics box, such as a phone or a tablet, or that shirt that’s wrapped in cellophane, or a football. Manipulating all these with the same system is very challenging and that’s a problem we believe we can solve.
But in that process, we have found that there are a lot of things that people would like us to focus on. Textiles is a good example, produce is a good example. So to bring it back to your original question, we are focusing on single markets at a time, but as we have focused on making a difference in those, that series processing if you will, we’re learning more that’s allowing us to open up new conversations in new markets.
Andra: What do you think is the single biggest obstacle that you’re facing in terms of startup growth? This could range from recruitment, to tools, to facilities, to funding, to things that I haven’t even thought of mentioning …
Carl: If I were giving advice to someone who’s going to start a robotics startup today, I would have them really spend a lot time on how to fund their company. We were at the Massachusetts Innovation UnConference two weeks ago, where we actually ran a session on funding the hardware startup. Traditional backers, the venture capitalists of the world, would really like to see traction.
Traction is the big word you hear in startups today. That means you have to have a prototype or you have to be running live pilots with potential paying customers or potential business partners. [You need to understand] the path and the funding necessary to get you from a really good idea or core technology, to where you have a pilot running and a customer, so you could stand in front of a venture capitalist and say, “This is real technology, we have solved the scientific risk. We can make it work and we have meaningful customers who are interested in the technology.”
By understanding that that’s the journey you’re going to be required to walk, I think you’re bringing clarity to [your] business plan. Unfortunately the world where you can have really great intellectual property or a great idea and get funding, it’s just becoming more difficult in recent years. So getting to that first customer contact is very, very important.
Andra: I agree, and I think by definition there’s a lot of intellectual property out there, but it’s hard to prove that it has value.
Andra: Unless you have developed traction. I’d actually avoided using that word even though it’s written down in my notes, so thank you for bringing that up. [Laughs.]
Carl: You’re welcome!
Andra: Is there anything that you would like to say that we haven’t covered? Do you have any other tips for startups that you’d like to share?
There are three things about doing a robotics startup today that I think are very important. One is, everyone is fascinated by robotics, and so whenever I’m out somewhere and someone hears the word robotics, they’re very interested in what we’re doing.
But as we’ve discussed many times at tradeshows, the public perception of robotics is very different from the reality of robotics, and [so second, you need to help] bridge that gap [by showing] the problems we’re trying to solve today versus the reality.
And third is that when you speak with investors – whether they are potential angel investors, or venture capitalists as investors – you need to help them understand the reality of robotics today, and also the vision of what robotics can be, and how you’re going to help move the industry forward. This is really the key I would say.
If you could remember those three things, you could be very successful in this industry.
Andra: That’s great. Thank you very much for your time. I really enjoyed the discussion!
All interviews are edited with great care for clarity of content, and may not represent the text of the original interview exactly. We can not assume responsibility for their accuracy.
If you liked this article, you may also be interested in: