Fruit flies serve as a key source of inspiration for robotics

On a black screen, a highly magnified image of a fly appears in white, walking steadily across a spherical surface on its six legs. “Just wait, it’s about to do the moonwalk,” someone says. We’re inside the EPFL Neuroengineering Laboratory, home to the Firmenich Next-Generation Chair in Neurobiology, led by Pavan Ramdya. Postdoctoral researcher Maite Azcorra uses a technique called optogenetics—shining precise laser pulses on the fly to activate targeted neurons. Almost instantly, the fly steps backward, resembling a dance move.

Since 2017, Ramdya’s 14-member research team has been investigating the nervous system of these tiny, two-millimeter insects. “Maite is focusing on how descending neurons from the brain regulate movement,” Ramdya explains. Their long-term goal is to reverse-engineer the fly’s brain and create robotic models based on it. A key advancement was building a digital twin to precisely simulate fly behavior, alongside a major discovery about how neural networks convert brain activity into coordinated motion. We settle into the office of the New York-born neuroscientist to delve into his research.

Can you explain the overall concept of your research program?

For centuries, humans have aimed to create machines that replicate the behavior of animals or people. Even in Ancient Greece, there were automated puppets—basic devices that mimicked human movement, serving as early examples of biomimicry. Our work follows that same principle, but with far more sophisticated tools and technologies that allow us to closely replicate the behavior and movements of creatures like the fruit fly.

Why did you choose to study Drosophila melanogaster in particular?

While more complex animals like mammals exist, they are much more challenging to study. On the other end, simpler organisms like C. elegans—a worm with only about 300 neurons—don’t provide as much insight into behavior. In contrast, fruit flies have around 100,000 neurons and exhibit a wider range of actions. Unlike worms, flies have legs they use to walk, groom themselves, and interact with their environment. For research in robotics and neuroprosthetics, it’s especially valuable to understand how an organism that uses both wings and legs functions. Flies strike the ideal balance: they’re simple enough for detailed study, yet complex enough to reveal meaningful information.

In your recent TEDx talk, you mentioned that future robots designed for space exploration might resemble fruit flies. Can you elaborate?

Absolutely. Robots sent to explore or even colonize other planets will need to operate independently, making quick decisions and navigating unpredictable, often hostile terrain. Although engineers have been striving to develop such machines for years, current robots still lack the agility and adaptability of a fruit fly. These tiny insects are remarkably capable—they can fly, maintain impressive stability with their six legs, and maneuver in three dimensions while using their legs for other tasks. Their versatility makes them a powerful source of inspiration for robotics design.

How could your research impact the fields of robotics and artificial intelligence?

While many engineers are focused on improving the physical components of robots—like motors and batteries—our work is centered on developing the control systems. Specifically, we’re interested in how a robotic fly might coordinate its limb movements. That’s why we study the nervous system of the fruit fly: to uncover principles we can use to design neural networks for use in robotics and AI. It’s also worth noting that these control systems aren’t limited to small robots—they can be scaled to fit machines of any size, even something as large as a house (though that might be a bit intimidating!).

Your research also explores other areas, doesn’t it?

Exactly. One fascinating feature of flies is that their legs are lined with mechanical sensors. We’re interested in understanding how they process all this sensory information to perceive their surroundings and detect nearby objects. For example, how do they decide when to lift a leg—or several—over an obstacle? To explore these questions, we’re working on creating materials inspired by the structure of the fly’s cuticle, embedding sensors into them for use in robotic systems.

Many experts in robotics and AI believe that for machines to truly learn, they need physical bodies that allow them to move and interact with their environment. Do you agree?

Absolutely. This is a core idea in both neurobiology and behavioral science—and it should be just as central in AI research. Animals are far more adaptable in their behavior than current robots. Machine learning engineers often highlight how human infants learn by actively moving, touching, and exploring their surroundings. That hands-on experience teaches them far more effectively than simply watching videos. The sensory systems I mentioned earlier—like those found on flies—play a similar role in helping organisms learn through direct interaction with their environment.

What are the main challenges in building systems that learn through exploration?

A major challenge is developing algorithms capable of interpreting sensory input. Without the ability to make sense of this data, it’s extremely difficult for machines to learn appropriate behaviors. The key point is that nature already holds the solution—it's embedded in the nervous systems of animals. That’s exactly what we’re aiming to understand. Rather than spending years building these systems from the ground up, why not learn from the efficient designs that already exist in creatures like flies?

Will this approach necessarily be quicker and more efficient?

Not necessarily. In reality, we’ll likely need a blend of different strategies. Animals have many biological functions and goals that don’t apply to robots—for example, reproduction or digestion. That’s why collaboration with biologists is crucial. They help identify which aspects of an organism are essential for our research and which can be disregarded, like neurons involved in waste elimination, so engineers don’t waste time on irrelevant systems. Our work truly depends on an interdisciplinary approach, combining expertise from both biology and engineering.

Is your ultimate goal to map the human brain?

To be honest, no—not for me personally. If I’m lucky, I’ve got about 40 more years, and I’d like to spend that time witnessing real breakthroughs in understanding how biological systems function. That feels achievable with the fruit fly, but mapping the human brain is a much more complex challenge. Maybe it’s just a matter of scale—perhaps if we multiplied the fly’s brain a million times over, we’d get something intelligent, and that would certainly be fascinating. But I’m not convinced it would reflect human intelligence. I don’t think the same approach we use for flies would work for humans—it would just take too long.

How does your approach to neuroscience differ from that of other neuroscientists?

In neuroscience, I’d say that over 99% of research focuses on human health and medicine. Most studies exploring how neuroscience can aid in disease treatment are conducted using mice or rats, since they’re mammals like us. Our research group offers two distinct perspectives. First, we explore neuroscience not only in terms of human health but also how it can be applied in robotics to design innovative machines. Second, we shine a spotlight on the smaller group of neuroscientists studying insects. Many insect species, like bees, are crucial to ecosystems—particularly in pollination—and are under threat. While fruit flies aren’t endangered, studying them can help us better understand at-risk species, which can contribute to conservation efforts. This perspective encourages a broader, ecosystem-focused view of the world, highlighting the importance of biodiversity.