Interview Transcript
Q1: If we could start with your name and where you were born, and just if you could tell us a little bit about your early education?
Antonio Bicchi: Yeah. I am Antonio Bicchi, and I come from Tuscany. My family’s from Siena. I was born in the northern part of Tuscany. And I now live in Massa there, which is a city that is sort of in the middle between Pisa—that’s where my university is—and Genova, where the Italian Institute of Technology is. That is my second affiliation and where I do much of my research.
As for early education, I grew up in a smaller city, a provincial town, and I’ve been living there for most of my life. I think I was lucky to do so. The education was not in a big institution, but accurate, I would say. Then I had classical studies, so I studied Latin, Greek, and all that.
Then I moved to engineering after a little thinking, because I was not sure. I might have taken a completely different path, but then I flipped the coin, and it was engineering rather than history of art. Having classical studies is interesting because mathematics is okay, it’s very close to Latin, but some other things were more difficult—more applied things like physics, for instance, is quite different in mindset.
Then I finished engineering at the University of Pisa, and I undertook a PhD at the University of Bologna in robot hands. Even during my master’s degree thesis, I did robot hands, so that is where I am rooted. This is what I have been doing for all my lifetime, together with other things, but that was one of the motivations for doing that connected to classical studies. I was studying the Greek philosophers, and I was interested in the debate that was there between two classical philosophers: Anaxagoras and Aristotle.
Anaxagoras was stating that humans are the most intelligent of animals because they have hands, and Aristotle was arguing against that and saying, “No, it’s because we are the most intelligent that we have hands”—the other way around. Whomever was right, we don’t know. We tend to be more in accordance with Anaxagoras, more causal. But the connection between hands and intelligence has been fascinating me for a lifetime.
Q2: How did you get first introduced to robotics? What was the first robot you were aware of?
Antonio Bicchi: I come from control theory, so I like the mathematical approach, the principled approach to engineering. When I met robotics, robotics was very, very young. I think at that time in Italy, there were maybe 4 or 5 groups, 4 or 5 professors doing robotics. I met a couple of them, and I considered that it was cool because it was somewhere where you could use your mathematical skills for understanding something that was completely new and not done before.
I thought it was an opportunity for understanding, doing, and merging the two parts that make things interesting to me—understanding and doing: understand to do and do to understand. That is what robotics is.
I’m very happy that I was in robotics because it’s really a melting pot. If you come from any of the components, then you have to stick to a methodology; you have a coherent set of colleagues that speak the same language. But in robotics, we can talk with psychologists, historians, neuroscientists, engineers, physicists, everybody. It’s a wonderful opportunity, so that reinforced me a lot in doing robotics.
Q3: In your journey, you mentioned in your master’s, even when you were working on robot hands, were you working theoretically or did you have actual robot hands that you worked with? What kinds of problems were you solving at the time?
Antonio Bicchi: We built a robot finger, not quite a hand, but we built a robot finger at that time in my master’s degree thesis. It was mainly an analytical thesis, but there was this building part that was posing problems. For us—people like me, and in our community there are many—the perfect loop is when you have a problem that comes from reality, and you look at it and say, “To solve this, I need tools that are more abstract.” You develop the abstract tools, apply them back to the original problem, and it works. That is an epiphany for us, and maybe you can also solve other problems with them.
Having the implementation and hardware together with the analytical tools and methods to solve them—robotics is just that: hardware and software, or hardware and ideas.
The very first problem was following with tactile feedback, following the contour of objects without leaving the object, and without pressing too much or too little. So being able to follow and reconstruct the shape of objects, which seems trivial today, but we were lucky in those days because we had problems that were outstanding and turned out to be feasible with the means that we had. Today, those problems are solved, so people have harder problems to solve.
Q4: Going from your finger, what was the next thing that you ended up working on?
Antonio Bicchi: From the finger, I moved to still tactile sensing. In my PhD thesis, I did several things. One was the development of a force-torque-based tactile sensor that, instead of using a surface material with several cells to detect contact, used an inside force-torque sensor and algorithms to find contact force, contact normals, tangential forces, detect friction, and predict and prevent slippage.
I was lucky to spend part of my PhD with Ken Salisbury at MIT, and together we made some advances there. Another part of my PhD was about adaptive grasping, again predicting and preventing slippage—understanding the closure of grasping when a hand is in firm control of an object, and where and how you should control the hand to keep that firm hold. So my thesis was about tactile sensing, grasping, and in-hand manipulation to some extent.
Q5: Did you continue to build your own hands along with working on the algorithmic side, or did you start anything new?
Antonio Bicchi: During my PhD, I didn’t have a hand myself. Later on, I started developing hands. Interestingly, the hands that we started developing were not anthropomorphic. For a certain time, we were interested in hands that could roll objects to manipulate them, so we built hands that exploited non-holonomic phenomena in rolling to reorient and manipulate objects. That was a beautiful mathematical problem, with an impact that was moderate in real life. We built the hand and did the experiments, but it was a bit too complicated to get into real life.
On the other hand, we also developed some multi-fingered, multi-articulated hands with motors in every joint—not very different from what we see today, except bigger, more cumbersome, and less modern.
I was frustrated, I must say. I graduated in 1988, and around the year 2000, I wrote a survey. In writing this survey, I realized, “We have done this, this, and that, but there are no real hands in the real world.” There were no hands in industry and no robotic hands in prosthetics. The hands present in prosthetics were rather simple or underused. So I took a sabbatical from hands for roughly ten years and did something else.
After about ten years, I encountered some neuroscience results. There was a lot of buzz about the work that Marco Santello, a young professor at Arizona State, was doing in neuroscience about hand synergies—the synergies that humans use to control motion of the hand.
I listened to a talk presenting these ideas. Meeting with neuroscience was very important because the problem with hands at that time was not the technology so much as the fact that it was too complicated to use. If you think of a hand, we have roughly 19 or 20 joints. Making sense of all these motions together takes a long time for a skilled programmer to write code, and then you’ve only programmed one little thing. That, together with the lack of robustness, was preventing hands from being applied in meaningful industrial applications.
More surprisingly, if you have a hand that has that many degrees of freedom and you want to use it as a prosthesis, you can forget about it because people who use prostheses have to control them through a very simple communication channel, which is most often just an open/close command that comes from EMG or even a body-powered harness. Having hands with 19, 10, 5, or even 2 degrees of freedom is too many.
That made us think that there must be a simpler way of controlling the hand so that I can talk with you while I do things with my hands without actively thinking about my hands. That was what was in those neuroscience stories. There are structures in the brain that distribute intent to a multiplicity of muscles. The hand has 350,000 nerve fibers communicating between the hand and the central nervous system. How can we have a chance to do this in technology?
The idea of synergies acts as funnels where many pieces of information converge to a few, or a few commands diverge to a multiplicity of actuators or muscles, making abstraction possible. We started thinking about that, and with these ideas, we made advances in the design of hands. We designed a hand that still has all the possibilities of motion of a human hand—19 or 20 different degrees of freedom—meaning a 19-dimensional space, but with only one input control. The control is such that the hand can move in a power grasp fashion, but the synergy architecture makes it possible for the hand to adapt the grasp to different objects.
Q6: How did these ideas extend into sensing, haptics, and prosthetics?
Antonio Bicchi: The same concept of funnels concerns sensing. Sensing is very important to hands. Out of the 350,000 nerve fibers, around 300,000 are for afferent signals—signals that go from the hand to the central nervous system bringing back feedback from sensors—and only 50,000 talk to the efferent system controlling the muscles.
Replicating sensing is very difficult, and again, synergies were useful to understand the higher-level percepts that organize this huge complexity. One idea in the science of touch was to use machinery developed for vision. Vision has been studied much more than touch—for instance, optical flow and the fact that we don’t necessarily look at things separately, but see how things move around to judge speed or time to contact.
We proposed having “tactile flow” instead of optic flow. Tactile flow is mapped in regions of the brain similar to where optic flow is mapped. The encoding of tactile flow relates to how we perceive the softness or hardness of an object—specifically, how quickly deformation spreads on our fingertips when pressing an object.
That intuition led to developing haptic displays that provide a sense of touching a soft or hard object without needing to stimulate every tiny sensing area in the fingertip. That is now used in haptic displays in prosthetics.
Prosthetics is a wonderful arena to work in because you work with robotics that directly talk to humans, and you have to make it work so that non-engineers can really use it. Another big lesson from prosthetics is that a person using robotic parts must consider them part of their body—giving commands intuitively and receiving credible feedback. That is the basis of what we call human-robot integration, which is the future.
We now look at the humanoids we have developed as whole-body prostheses. It’s not just the hand or the arm; it’s your own body, but you still give commands and receive feedback without being bothered by minor details. You don’t want a robotic hand that grabs a can every time it sees one—you only want to grab it when you are thirsty. The real point today is giving the person the ability to use the robot as an integrated part of their body.
Q7: When you’re thinking of humanoids in this fashion, are you seeing them as more teleoperated in a sense, even if they’re much more naturalistically teleoperated than what we have nowadays? Or do you think of them as separate from humans?
Antonio Bicchi: Absolutely, that’s our point of view—that’s what we are interested in. I’m not saying there is no use for autonomous humanoids somewhere; we will see whether they go on production lines in industries and replace workers. I’m not a big believer, but we will see.
My perspective is rather that I want to have an avatar—a second copy of me or a prosthesis, if you want. I want to be able to have one at home to tend to my pet, do the work I’m not able to do, or be with my mom who is aging up to help. I don’t need to be in charge of it every second; I can teleoperate very metaphorically, but I must be able to tell it exactly what to do, and it should do what I want it to do rather than doing things on its own. Complete autonomy doesn’t necessarily mean helpful, and it might end up having a will of its own in some sense.
Q8: In all of this work inspired by findings from neuropsychology, were you collaborating with scientists from other disciplines? Can you tell us a little bit more about that?
Antonio Bicchi: Yes, I’ve been privileged to work with many colleagues from different disciplines. This is in part due to the nature of research funding we had in Europe, where it was very possible to have grants tackling ambitious problems with partners across psychology, robotics, mathematics, philosophy, and neuroscience.
Today, worldwide, there are shorter return-on-investment requirements asking for finalized applications. I don’t think that’s completely bad, because as engineers, our job is to find concrete problems, abstract them to solve them, and return to concrete solutions.
Additionally, schemes like the European Research Council (ERC) give very open-minded, generous, and free grants. Many of these ideas were possible because of that freedom. For instance, in a recent project, we worked together with a neuroengineer, Professor Dario Farina from Imperial College (who reconstructs spiking motor neuron signals in the spine from EMG signals in an amputee’s stump), and Professor Oskar Aszmann, a neurosurgeon in Vienna (who does targeted muscle reinnervation). He rewires the nerves that used to serve the missing hand to active muscles in the stump, so we can access the bio-hub input/output ports. We connected those by understanding the language spoken by the spiking motor neurons to talk directly to our prosthetic hand.
Q9: Have you had opportunities to see people using your prosthetic devices in real life?
Antonio Bicchi: Definitely, many times. We recently wrapped up a two-year project where our hands were tested at home by tens of patients in the United States every day for a month, taking measurements on usage frequency, challenges, and advantages. We’ve tried them on hundreds of amputees now.
Most appreciate it because it’s different from other hands—it’s soft. That’s another key lesson from prosthetics: the need for soft robotics. Beyond robustness and safety, softness is essential for social interaction—shaking hands, hugging, or touching your face. You wouldn’t use a rigid, engineered hand to touch your face, so softness is crucial for human compatibility.
Q10: Did this need for soft robotics in prosthetics also guide your broader transition into soft robotics research?
Antonio Bicchi: Yes, though I’ve been aware of soft materials for a long time. The group I started working with under Paolo Dario in the 80s was already considering soft materials. Later, we worked on softness for robot arms in collaborative robotics (cobots) in the early 2000s.
Prosthetics and neuroscientific understanding reconnected that softness back to hands. Human hands are soft—we can deform them. That softness provides adaptiveness, social compatibility, and durability; rigid hands break if subjected to the everyday impacts our hands absorb naturally.
Q11: What did you explore during your ten-year sabbatical from robot hands?
Antonio Bicchi: From 2000 to 2010, I worked on human-robot cooperation and safe interactions between humans and robots. Together with others, we developed concepts of safety and how to make robots inherently safe, designing new robots and assessing impact safety.
In a project with DLR in Germany, we performed the first quantitative risk assessment of a robot hitting a human to codify what was dangerous versus safe. Having those standards codified and standardized made it possible for industries to deploy cobots alongside humans, because compliance enabled insurance coverage. Now, cobots make up around 20% of the industrial robotics business, and those early safety projects were instrumental in favoring that shift.
Q12: Looking back at that sabbatical, what findings or achievements were you most excited about?
Antonio Bicchi: Designing new control methods and robots that made it clear what was dangerous versus what was okay. Quantifying and standardizing those risk levels made it possible for industries and regulations to allow humans and robots to work side-by-side safely.
Q13: You were one of the founding Editors-in-Chief of IEEE Robotics and Automation Letters (RA-L). How did that come about?
Antonio Bicchi: At the time, robotics had selective, high-quality conference papers that were well-cited, but they didn’t count as journal publications, which created practical administrative issues for researchers going for tenure. Additionally, publication processes were slow.
The core innovation of RA-L was creating a unified submission process where a paper could be reviewed for presentation at a conference while simultaneously being evaluated for journal publication. After addressing initial misunderstandings alongside colleagues like Stefano Stramigioli and Alessandro De Luca, it gained strong support and unanimous approval within IEEE.
I ran it for three years after launch, and under subsequent Editors-in-Chief like Alison Okamura and Tamim Asfour, the journal has grown rapidly and serves a huge community today.
Q14: Looking ahead, what do you see as the future of robotics that you are most interested in studying?
Antonio Bicchi: The future we are interested in studying is human-robot integration—creating a dynamical system where both the human and the robot participate together. The robot observes an action initiated by the human, infers intention from that initial motion, and completes the action.
This paradigm extends beyond prosthetics to teleoperated avatars. You can control a robot in another room by simply initiating a motion; the robot infers your intention and completes the task so you can relax. Recently, we’ve used these avatar systems with individuals who have spinal cord injuries to help them reach and perform tasks they otherwise couldn’t.
Q15: What advice or message would you share with students or early-career researchers interested in getting into robotics?
Antonio Bicchi: Enjoy it! Robotics is a lot of fun. Enjoy the freedom robotics gives you, the diversity of people you meet and learn from, and the fact that you can bridge theory and physical experiments. Doing what brings you enjoyment is the key to going to bed happy about what you’ve accomplished during the day. In robotics, doing good work and having fun go hand in hand.