In this interview, Fabio Bonomo, CEO of qbrobotics, explains why durability, not precision, remains the hardest constraint in robotic hands, and why mechanical compliance matters only when it can survive industrial use over millions of cycles.
qbrobotics emerged from the University of Pisa and IIT research ecosystem. In your experience, what is the single hardest step when translating variable stiffness theory into an industrial product that must run millions of cycles without failure?
There isn’t a single most difficult technical step. the real challenge is a shift in mindset. Moving from theory to an industrial product means designing the entire system — mechanics, electronics, and control — to operate for millions of cycles under real-world conditions. In this transition, aspects such as friction, wear, thermal management, and software robustness become central. The difficulty does not lie in a single component, but in ensuring that the entire system remains reliable and stable over time.
Many robotics companies are investing heavily in vision models and AI-driven grasp planning. Where do you believe software intelligence alone reaches its limits without mechanical compliance?
Control, without an effective and efficient gripping system, will never be able to achieve a truly robust and adaptive grasp. Likewise, a mechanically compliant gripper without an adequate vision, recognition, and object targeting system will not be able to interact reliably with the external environment. Software intelligence alone reaches its limits when mechanical compliance is missing — compliance that can absorb uncertainties, estimation errors, and variability in the physical environment. It is precisely the right combination of artificial intelligence and mechanical intelligence that, at this historical moment, will allow robotics to take a real leap forward and make systems increasingly advanced.
Traditional industrial robotics optimized for precision and repeatability. Do you see a structural shift toward compliance and adaptability, and what practical evidence convinced you this shift is real rather than theoretical?
There are still many tasks where precision and repeatability remain fundamental. This is therefore not a replacement of traditional robotics, but an expansion of its field of application. The shift toward compliance and adaptability is real because it enables access to a range of tasks that were previously difficult to address or economically unsustainable with rigid robotics. In contexts where traditional robotics struggles — such as handling delicate, variable, or imperfectly positioned objects — solutions based on soft robotics and compliant actuation are demonstrating clear advantages. Application cases such as food processing, agritech, and certain luxury segments show how more adaptive systems can manage deformable, fragile, or highly variable objects more effectively.
Can you share a specific deployment where intrinsic mechanical compliance solved a problem that sensor-based safety systems could not?
A “simple” example concerns wear testing of a handbag or clutch in the luxury sector. In these cases, the product is subjected to repeated grasping cycles to simulate long-term use and assess how the material wears over time. A rigid system, even if equipped with force sensors, tends to apply localized and unnatural pressures that are difficult to modulate realistically. A hand with intrinsic mechanical compliance, instead, adapts to the object’s geometry and distributes pressure more evenly, replicating a soft and adaptive human grasp. In this scenario, compliance does not act through control, but physically and immediately, compensating for shape variations and absorbing micro-differences that emerge during interaction.
qbrobotics chose to integrate into ecosystems like UR+ rather than build a full robot stack. What advantages and constraints come with that platform strategy?
The decision to integrate into ecosystems such as UR+, Dr. Dart Suite, CRX, etc., stems from a substantial technical and commercial difference between developing a robotic arm and designing an end-effector. These involve different skills, regulations, and specializations; even UNI EN ISO standards clearly distinguish between robotic arms and gripping devices. From a market perspective, robot manufacturers already have established distribution channels, service networks, and installed bases. Integrating into these ecosystems allows us to leverage our specialization by bringing interoperable intelligent grippers onto widely adopted platforms, instead of competing with a new anthropomorphic arm in a crowded sector.
The advantage is focusing on innovation in gripping technologies and addressing problems that have not yet been tackled, or have only been handled in traditional ways. The constraints are linked not only to regulatory compliance, but also to mechanical sizing requirements for wrist flange interfacing, as well as software protocols, power supply voltages, pinouts, and technical specifications defined by different robot manufacturers.
Variable Stiffness Actuators add mechanical complexity. What trade-offs did you accept in order to make VSA commercially viable at scale?
Variable Stiffness Actuators introduce greater mechanical complexity; the key trade-off is therefore between maximum theoretical performance and industrial sustainability. We prioritized simpler, more robust, and more easily industrializable architectures, reducing critical components and sensitive tolerances, and focusing on modularity and scalability. This sometimes means not pushing every performance parameter to its extreme, in favor of reliability, manufacturing repeatability, and sustainable costs.
Soft robotics has historically struggled to move beyond research labs. What did you change internally to ensure qbrobotics became a repeatable industrial supplier rather than a research-driven company?
To move beyond a purely academic logic, we worked on two levels: technical and cultural. From a technical perspective, we redesigned components and processes to ensure repeatability, robustness, and compliance with certification standards. No longer prototypes that work in the lab, but products designed to be industrialized, mass-produced, and maintained over time. Even more important was the mindset shift: moving from a research-oriented approach focused on technological demonstration to a market-oriented approach. This means designing with precision, reliability, and scalability in mind, with the goal of solving concrete application cases and generating value for customers. In short, not only technological innovation, but industrial discipline and a focus on real-world applications.
In one moment of product iteration that did not go as planned, what did you learn about the gap between technical elegance and real customer demand?
We learned that technical elegance does not always align with what the market is willing to pay for or adopt. In some cases, we realized we were solving a problem in an extremely sophisticated way that could have been addressed with a simpler and sufficiently effective solution. This led us to simplify the solution, reducing complexity and costs without compromising application value.
The main lesson was to find a balance between engineering refinement and real utility: not designing the most technically elegant solution in absolute terms, but the one most aligned with concrete needs, ease of integration, and economic sustainability for the customer.
As humanoid robotics gains attention, many prototypes still struggle with durability. From your perspective, what is the most underestimated engineering challenge in robotic hands today?
From qbrobotics’ perspective, the most underestimated engineering challenge in humanoid robotic hands today is combining naturalness, softness, and dexterity with structural stiffness and industrial durability. It is possible to design a very soft and adaptive hand for laboratory demonstrations, just as it is possible to create an extremely rigid and robust hand for repetitive tasks.
The real difficulty lies in integrating these two worlds: ensuring natural compliance capable of interacting safely and intuitively with objects and people, without compromising mechanical stability, precision, and system lifespan. A truly effective hand must be soft when interacting with the environment — especially with people — but become sufficiently robust when applying force or maintaining accuracy.
It must also do so for millions of cycles, without softness translating into premature wear, loss of calibration, or structural failure. In other words, it is not only about replicating the form or degrees of freedom of the human hand, but about reproducing its balance between adaptability, dexterity, and solidity. It is precisely in this integration of intrinsic compliance, dexterity, intelligent control, and mechanical robustness that the true maturity of robotic hands is measured.
Mechanical compliance can absorb uncertainty during physical interaction. Do you believe this property meaningfully reduces the burden on AI training or reinforcement learning systems in real deployments?
Mechanical compliance can certainly absorb part of the physical uncertainty during interaction, reducing the need for continuous compensation by control systems. In this sense, it eases the problem by making the system intrinsically more tolerant to positioning errors, shape variations, or micro-impacts.
However, I do not believe it significantly reduces the role of AI or reinforcement learning. Rather, it makes them more effective. Artificial intelligence has a strong multiplicative effect when operating on a mechanically well-designed platform. If the mechanics can already absorb part of the uncertainty, AI can focus on manipulation strategies, adaptation, and optimization instead of compensating for rigidity or structural instability.
In other words, compliance does not replace AI: it creates the conditions for AI to work better, with less data, less destructive exploration, and greater robustness in the real world. It is the combination of mechanical intelligence and algorithmic intelligence that enables scalable and reliable applications.
Europe has strong mechanical engineering traditions but a different capital environment than Silicon Valley. How does that shape qbrobotics’ long-term competitive positioning?
Europe offers a strong tradition in engineering and research, which supports deep and structured product innovation. Even in a capital environment that is less aggressive than Silicon Valley, there is a solid digital and industrial vocation. For qbrobotics, this means focusing on concrete product innovation: robust solutions engineered for real-world use, with strong technological and application-driven content. It is perhaps a less growth-at-all-costs approach, but one centered on distinctive, reliable, and sustainable products over the long term.
Looking ahead to 2030, what assumption about human–robot collaboration do you think the industry will reconsider, and where will flexible actuation play a decisive role?
Looking toward 2030, I believe the industry will reconsider an implicit assumption that has guided much of collaborative robotics so far: the idea that safety in human–robot interaction must primarily be ensured through force limitations, speed restrictions, and software control.
More and more applications will require robots that are not only “limited,” but intrinsically safe and physically compatible with humans. Here, flexible actuation will play a decisive role. Mechanics capable of absorbing energy, adapting, and naturally modulating stiffness allow for smoother, less artificial, and more productive interaction without sacrificing performance.
By 2030, it will not be enough to have robots that stop when contact is detected; systems will need to share space and action with humans in a continuous, intuitive, and safe way. In this scenario, compliance and variable stiffness will not be optional technologies, but enabling architectures for truly integrated human–machine collaboration.
Editor’s Note
This interview examines a broader shift in robotic manipulation from rigid precision toward compliant, durable systems that can operate reliably in real industrial environments.
