Interview Transcript
Q1: Hi, professor. Thank you so much for coming today. We start with your full name and date of birth and place of birth.
Paolo Fiorini: My name is Paolo Fiorini. I was born in Verona, Italy, on February 1st, 1953.
Q2: Could you tell us a bit about your family history and background?
Paolo Fiorini: Well, my father was an engineer, so I always wanted to build things. When it was time to graduate in the early-to-mid ’70s—in ’76 when I graduated—it was the beginning of microprocessors. It was the era of the 8080, 8085, 8084, and 8088. So I started doing my master’s thesis on developing control systems using microprocessors for different types of devices. At the university, I developed some language processors for microprocessors.
Then I immediately found a position in the industry because it was a very hot topic, and companies were looking for that expertise. My first job was developing computerized domestic appliances. We developed the first computerized washing machine and a network for kitchen ovens. Later, I moved on to networks for public transportation and networks for industry. I was working for a company that made appliances and industrial cooking systems, like industrial stoves. That was actually my only foray into consumer products. It was in Italy from 1977 to 1980, in a city called Pordenone.
At that time, there was a large manufacturer of domestic appliances that had just opened a research center. It was a fantastic experience because they attracted Italian emigrants back to Italy from the United States, France, and Germany. There was a very international and multidisciplinary group that was really fantastic to work with. That started my interest in eventually going to the United States and doing further work, including my PhD.
Q3: Did you do a PhD in Italy?
Paolo Fiorini: At that time, there was no PhD program in Italy. So I did my master’s, which was sufficient back then. I did my PhD much later when I was in the United States.
Q4: Do you remember the time when you got specifically interested in robotics?
Paolo Fiorini: Yes, of course. After my work in that company—it doesn’t exist anymore, so I can say the name; it was called Zanussi—I moved to another company in Milan. From that company, they sent me on a consulting job to California, in Anaheim Hills of all places. At that time, in 1980, the difference between Italian academia and American academia was incredible—it was a different world.
The people working with me on this job suggested I visit the nearby university, which was close to the University of California, Irvine. I was able to get a one-year leave to pursue my master’s degree. I asked myself, “What is something that is really new that I don’t know anything about?” Robotics came up. In 1980, nobody talked about it yet; the first textbook on robotics didn’t come out until around 1982. It was really a frontier of research. So I started doing research in robotics, developing a dynamic model of a small robotic arm.
Q5: Growing up in Italy, were the first robots you heard about from comic books, popular culture, or anything like that?
Paolo Fiorini: I was actually working on early robots. A few years after finishing my master’s, in 1985, JPL—the Jet Propulsion Laboratory—was forming a robotics group and hiring. I applied, and one of the managers had seen my thesis and publications, so they hired me. I was able to work on those first robots. The workhorse of our laboratory was the PUMA 560.
At that time, robotics wasn’t part of popular culture yet—there were no movies or cartoons about it. But it was being used for space exploration and space missions. So we were able to develop some of the first robotic applications for space.
Q6: What were the main research questions you were pursuing back then?
Paolo Fiorini: That’s a very important point. At that time, robotics was really in its infancy, so robots were not able to do much more than just move. You had to be very careful because motions could be very fast and risky. The main research question was how to make sure you could teleoperate a robot, effectively providing human intelligence to the machine.
We were interested in basic research questions: How do you manipulate and grab objects? How do you perform assembly tasks? Again, the objective was space applications. At that time, space repairs were done by astronauts doing Extravehicular Activity (EVA), which was very dangerous and expensive. The goal was to enable teleoperation from Earth or from a space station to perform these repairs.
Teleoperation was the main focus of research. How do you ensure telepresence—making sure you can feel everything the robot is feeling in a remote environment using touch, vision, and force? Below telepresence, the core layer was force feedback, which is now called haptics. We were one of the first laboratories to develop a robot with full force feedback for surgery and space applications.
It is funny because some of those key questions are still not fully answered today. Force feedback is a very important topic that is just not receiving enough attention right now. Today, everything is about AI and learning-based approaches. But all the psychology behind force feedback—whether it is truly useful, whether it could confuse the operator, or what advantages it brings—is still not fully understood.
Q7: How have the main research questions changed over time?
Paolo Fiorini: They have changed for many reasons. As I mentioned, I didn’t have a doctorate when I arrived in Italy, but I was working in these very advanced research laboratories. Being one of the few without a doctorate, I said to myself, “It’s not possible, I need to get my doctorate.” So I decided to do my PhD at UCLA, complementing my activity in teleoperation.
Around 1986, robotic autonomy was really not considered by anybody. I did my PhD in robotic autonomy and developed an algorithm for autonomous car navigation that is still cited today and served as a foundation for autonomous vehicle navigation.
I took a two-pronged approach: on one side, teleoperation to keep the human involved; on the other side, determining how much autonomy we can give to the robot to support the operator. In teleoperation, communications can break. If you are doing a safety-critical operation, you want the robot to be able to complete it or reach a safe state autonomously. Thus, autonomy and teleoperation must coexist.
Q8: What were some of the big challenges you faced in your career—institutional, technical, or financial?
Paolo Fiorini: All of the above! It is very difficult to conduct research, make meaningful contributions, and keep things funded all at the same time. I was in a very fortunate situation because I received my professional training in the US, including my PhD and 15 years of experience at JPL. That formed a very unique professional culture.
Back then, we had a lot of freedom to explore and experiment. Today, my colleagues at JPL face very tight deadlines and high stress, but 40 years ago, it was much more relaxed. I was able to develop a strong desire to explore new avenues, and NASA was very supportive of that.
When I returned to Italy, this mindset really helped me secure significant funding from the European Union. I already had experience competing, writing proposals, and getting experienced teams to work together. The challenge was finding the equilibrium between all these elements: finding good people and students, securing funding, and generating good ideas. All three must coexist.
Q9: What tools came along during your career that made a major impact on your research and opened up new possibilities?
Paolo Fiorini: The primary breakthrough was the development of software infrastructure for robotics. At the beginning, you had to code everything from scratch to make the robot move. Now, you have software libraries for control, interaction, and motion planning.
The flip side is that some researchers, starting with my former students, don’t develop anything new from scratch anymore; they just search the internet for existing solutions. So while these tools allow us to accomplish more work thanks to existing infrastructure, they can also limit human creativity.
Now, the next major shift is AI taking over. I am really curious to see how AI generative models and mechanical engineering models will coexist and whether they will find a healthy balance.
Q10: Do you see a difference in how students today approach research compared to how you did it back then?
Paolo Fiorini: Absolutely. It starts with simple things like going to the physical library, which used to be a feast of discovery for me. You would look for a book, spot another title next to it on the shelf, and discover a whole new world. Today, people just search for a specific paper online, read it, and that’s it—there is less cross-fertilization of ideas.
Furthermore, when developing a project today, people tend to look for pre-existing solutions rather than inventing new ones, which can be limiting. This has contributed to a proliferation of papers where individual contributions are often small and incremental rather than fundamental leaps forward.
Q11: You mentioned being successful in getting European funding. Who was the main funder of your research?
Paolo Fiorini: Thanks to my professional background in the US, I was trained to compete and write proposals. In Europe, the main sponsor is the European Union through its Framework Programs. These programs cover the whole spectrum, from funding initial concepts to supporting company creation.
Because of my background in space robotics, the closest field to space was medical robotics. Both involve teleoperation, are safety-critical, and require high certification and precision. I returned to Italy in 2001, at the very beginning of robotic surgery. I was involved in explaining to clinicians what a surgical robot was and how it could be used.
We started with an engineering mindset, thinking the operating room was like a machine shop where you could just place a robot, position the patient, and perform the operation. Of course, it wasn’t that simple, but fortunately, the proposal reviewers didn’t realize that yet, so we got funded! Eventually, my later projects focused on autonomy and making robots intelligent enough to assist in surgery.
Q12: Have you founded companies as well?
Paolo Fiorini: Yes, I have started several companies. Three years ago, I retired from academia. In Europe, and especially in Italy, once you retire, you are largely cut off from university activities. However, I still felt I had more to contribute, so I started a company focused on a specific niche that I believe is very promising, though getting traction remains challenging.
Q13: Is there anything you would have liked to pursue in your career but didn’t get the chance to?
Paolo Fiorini: Yes, many things! When I was in the United States, I took advantage of the research environment in the laboratories, but I didn’t pay enough attention to business creation. For example, the algorithm I developed for my thesis could have been the foundation of a successful company.
Additionally, one of my last projects at JPL was a surgical robot designed for eye surgery. In eye surgery, providing physicians with a sense of touch is critical because it is a very delicate environment. That concept could have been turned into a commercial product.
My initial lack of business understanding led to mistakes when I later started companies without fully understanding the constraints and requirements of the commercial world. I recently launched a startup called Needle Automation Robotics. I was giving a talk at an innovation forum about three common misconceptions researchers have: thinking that having great technology, a validated application, and a working prototype tested on cadavers is enough. It isn’t. You need to thoroughly work out the economic and market aspects to convince investors. I wish I had learned those business lessons much earlier in my career.
Q14: What do you see as your most important contributions to robotics?
Paolo Fiorini: There are a couple. First is my thesis algorithm for robotic navigation in dynamic environments, known as the Velocity Obstacle method. Interestingly, a paper at this conference won a best paper award based on a variation of that algorithm, which shows it is still actively used and relevant.
Second, my last project at JPL focused on hopping robots, a concept currently being considered for lunar bases. In low gravity, jumping or flying is the most efficient way to move around. We developed small devices that could hop from place to place, offering an alternative to single large rovers by using a fleet of small hopping robots. Although it wasn’t funded at the time, the concept is being revisited today.
Finally, when I moved back to Italy, I helped pioneer robotic surgery in areas like surgical training, performance quantification, and what is now called “frugal robotic surgery”—building low-cost, high-performance medical devices. By leveraging AI and advanced algorithms, you can lower component hardware costs while maintaining high performance. The surgical machine we are developing aims to cost under 200,000 euros or dollars, compared to the 3 to 4 million dollars required for conventional surgical robots, making the technology far more accessible.
Q15: What timeline do you estimate for this lower-cost medical robotics model to become widespread?
Paolo Fiorini: Things are moving very fast. Within the next five years, we will see significant shifts. The current financial model of medical robotics is unsustainable because systems are too expensive and require too much training. We urgently need accessible solutions.
A small group of researchers in the UK, the Netherlands, and Italy are addressing this. For instance, there is a fantastic group at TU Delft in the Netherlands working on low-cost surgical tools and operating rooms. This is vital not just for developing countries, but also for wealthy nations where many regional hospitals lack advanced surgical care.
This leads back to teleoperation and telemedicine: pairing low-cost robotic systems with reliable communications allows for a distributed healthcare network that supports remote diagnosis and treatment everywhere at a fraction of current costs.
Q16: When introducing your robots into surgery, were there any unexpected effects or outcomes that surprised you?
Paolo Fiorini: Robots, at least in the traditional deterministic way I approached them, are very predictable, so unexpected surprises were minimized through careful preparation.
The real surprises will come with AI and machine learning. There is a concept that truly concerns me called “emergent misalignment.” Emergent behavior occurs when a machine develops unprompted actions; emergent misalignment happens when those behaviors become harmful or deviate completely from intended human values.
For instance, experiments have shown AI systems trained on imperfect examples not only learning bad habits, but fundamentally altering their operating parameters. In physical robotics, we have already seen accidents with industrial robots. These unexpected, unintended AI behaviors are the surprises we must work hard to prevent.
Q17: What would you recommend to students interested in entering the field of robotics today?
Paolo Fiorini: Robotics is a fantastic field! First, I would advise placing less immediate emphasis on AI alone and focusing more on algorithmic approaches, modeling, and underlying physical principles. Modeling allows you to abstract behaviors and truly understand how systems work, whereas AI often bypasses that understanding.
Second, look toward high-impact application areas like personal support, home care, and elder assistance. We have an aging society with fewer caregivers available, so assistive robots for the home are essential. I often give public lectures to elderly audiences and am continually impressed by their pragmatic openness to technology. If a solution works, they welcome it.
Designing home-assistive robots is complex because home environments are unstructured compared to factories. This is where AI and human-robot interaction can genuinely help interpret user intent and adapt to dynamic human needs. I highly recommend students explore that direction.
Q18: Is there anything else you would like to mention before we conclude?
Paolo Fiorini: I would emphasize the need to support the creation of new startups that bring research technologies to market. We need better frameworks to help academic researchers navigate entrepreneurship.
When researchers attempt to commercialize technology, they often feel like tiny goldfish swimming among sharks. Major universities sometimes provide business incubators, but generally, technical research and business development remain disconnected.
Organizationally, communities like IEEE RAS could provide more structured mentorship and commercial guidance so young researchers with brilliant ideas don’t feel isolated or risk having their innovations taken advantage of. Improving that bridge between academia and industry would make a major difference for the next generation.