A robot lets you control one variable at a time, run the same test a hundred times identically.
For centuries, the fluid grace of fish has inspired wonder and imitation, but understanding the deeper mechanics of how they swim and school has remained elusive to those who could only watch. Now, researchers are turning robotic fish — once conceived as engineering curiosities — into precise scientific instruments capable of intervening in, rather than merely observing, the living world. By placing adaptive machines directly into biological systems, scientists are beginning to decode the feedback loops that connect muscle, motion, sensation, and collective behavior. It is a quiet revolution in method: the artificial becoming a key to the natural.
- Traditional biology has hit a wall — living fish cannot be paused, reset, or surgically altered mid-experiment, leaving core questions about locomotion and schooling frustratingly out of reach.
- Robotic fish are now crossing from engineering showcases into genuine scientific tools, capable of measuring internal forces and energy use that remain invisible inside any living animal.
- A critical gap persists: most robot-fish research still serves engineering goals, and only a small fraction employs the closed-loop, adaptive systems needed to mirror how real fish actually sense and respond to their world.
- Closed-loop robots — those that react to their environment in real time — are proving essential for probing how fish decide to join a school, adjust their stroke, or follow a neighbor's lead.
- The field is now converging toward an integrated approach, weaving robotic experiments, computer simulations, and emerging measurement technologies into a single framework for understanding behavior across scales.
- The governing question has shifted from 'Can we build a robot that swims?' to 'What can a robot reveal about the principles evolution has encoded in fish?' — and that reframing is opening entirely new doors.
Fish are among nature's most accomplished movers, navigating water alone and in schools with an efficiency that has long fascinated both biologists and engineers. For years, researchers studied these behaviors through observation and inference. Now, a new generation is building artificial fish not to replicate nature, but to interrogate it.
Robotic fish have matured from engineering prototypes into genuine scientific instruments. Unlike cameras or sensors that observe from a distance, these robots can be placed inside the very system under study — interacting with real fish, responding to stimuli, and being modified in ways no living subject could tolerate. A researcher can alter a tail's shape, adjust fin-beat frequency, or change how the robot responds to sensory input, then measure the precise consequences. This kind of repeatable, controlled experimentation has been largely unavailable to fish biologists until now.
The advantages are real: identical experimental conditions run hundreds of times, direct measurement of internal forces and energy expenditure, and the ability to test specific hypotheses about sensory feedback and collective movement. Yet a survey of existing research reveals a telling gap. Most robotic fish studies remain focused on engineering — on building better robots — rather than on answering biological questions. And among those that do pursue biological insight, only a small fraction use closed-loop systems: robots that sense their environment and adapt in real time, as living fish do.
This distinction is crucial. Fish do not execute fixed swimming programs; they respond continuously to currents, neighbors, and obstacles. A robot locked into a predetermined motion can answer only limited questions. A closed-loop robot, by contrast, can model the feedback mechanisms that govern how a fish joins a school or adjusts its stroke in response to a neighbor — making it a far more powerful lens on behavior.
The emerging path forward is integrative. Researchers are combining robotic experiments with computer simulations and new measurement technologies to trace the feedback loops that run from muscle contraction through body motion to sensory perception to behavioral choice. No single method can illuminate all of these scales; robotics provides controlled intervention, simulation provides theoretical scaffolding, and new technologies offer unprecedented access to living systems in motion.
What is taking shape is a genuine partnership between engineering and biology — one in which the robot is no longer merely inspired by the fish, but becomes a collaborator in understanding it. The field is asking not whether a machine can swim, but what a machine can teach us about the principles evolution has spent millions of years refining.
Fish have always been masters of motion—they glide through water with an efficiency that engineers have spent decades trying to replicate. Their ability to swim alone and coordinate in schools represents some of nature's most elegant solutions to problems of physics and social organization. For years, biologists studied these behaviors by watching, measuring, and theorizing. But a new generation of researchers is taking a different approach: they're building artificial fish to ask questions that living animals cannot easily answer.
Robotic fish have emerged from the engineering labs where they were first conceived as design exercises and are now becoming genuine scientific instruments. Unlike a video camera or a microscope, which observe from the outside, these robots can be placed directly into the system being studied. They can interact with real fish, respond to stimuli in real time, and be modified in ways that would be impossible with living subjects. A researcher can change the shape of a tail, adjust the frequency of a fin beat, or alter the robot's response to sensory input—and then measure exactly what happens next. This kind of systematic experimentation is the foundation of science, yet it has been largely unavailable to those studying fish behavior.
The advantages are substantial and concrete. Robotic models allow researchers to run the same experiment dozens or hundreds of times with identical conditions, something no two living fish will ever provide. They can measure internal forces and energy expenditure that remain invisible in living animals. They enable direct testing of specific hypotheses about how sensory feedback shapes movement, or how individual fish decide to follow or lead in a school. Traditional biology often relies on observation and inference; robotic approaches permit direct intervention and measurement.
Yet a review of the existing literature reveals a curious gap. Most studies using robotic fish have focused primarily on engineering—on building better robots, refining designs, improving efficiency. While this work has value, it has contributed relatively little to fundamental biological understanding. The robots have been treated as ends in themselves rather than as tools for answering biological questions. Among the studies that do aim at biological insight, an even smaller subset employ what researchers call closed-loop systems: robots that sense their environment and adapt their behavior in real time, much as living fish do. These adaptive systems are precisely what's needed to probe the feedback mechanisms that govern how fish move and interact, yet they remain underrepresented in the literature.
The distinction matters because fish do not simply execute a predetermined swimming pattern. They respond to water currents, to the presence of other fish, to obstacles and opportunities in their environment. A robot that cannot adapt—that merely repeats a fixed motion—can answer some questions but not others. It cannot reveal how a fish decides to join a school, or how the presence of a neighbor changes the swimmer's own stroke. Closed-loop robots, by contrast, can be programmed to respond to sensory input just as a real fish does, making them far more powerful tools for understanding the mechanisms beneath behavior.
The path forward involves integration. Researchers are beginning to combine robotic experiments with computer simulations and new measurement technologies that can capture details of fish movement and physiology at scales previously out of reach. The goal is to understand the multi-scale feedback loops that connect muscle contraction to body motion to sensory perception to behavioral choice. These loops operate across levels of organization—from the molecular to the whole-animal to the group—and no single method can illuminate all of them. Robotics provides the controlled experimentation; simulation provides the theoretical framework; emerging technologies provide unprecedented access to the details of living systems.
What's emerging is a new partnership between engineering and biology. The robot is no longer merely inspired by the fish; it becomes a collaborator in understanding how the fish works. This shift from curiosity to utility, from engineering showcase to scientific instrument, represents a maturation of the field. The questions being asked are no longer "Can we build a robot that swims?" but rather "What can a robot teach us about the principles of swimming and social coordination that evolution has encoded in fish?" That reframing opens doors that observation alone cannot unlock.
Notable Quotes
Robotic fish are shifting from engineering curiosities to experimental partners in biology— Nature research perspective
The Hearth Conversation Another angle on the story
Why do we need robotic fish at all? Can't we just watch real fish and figure out how they work?
Watching tells you what happens, but not always why. If you want to know whether a fish changes its tail beat because of water pressure or because it sees a neighbor, you need to control one variable at a time. A robot lets you do that. You can change its tail shape and run the same test a hundred times with identical conditions. A real fish won't cooperate.
So the robot is like a controlled experiment in a lab coat?
Exactly. But with a twist—it can be placed in the water with real fish. It becomes part of the system. You can watch how actual fish respond to a robot that behaves in specific, repeatable ways. That's something you can't do by just filming.
I'm hearing that most robot-fish research hasn't actually helped biology much. Why build them if they're not answering biological questions?
Most of the work has been engineering-focused—proving you can build a better robot, make it more efficient, refine the design. That's valuable for robotics, but it doesn't necessarily teach us about fish. The real power emerges when you use the robot as a tool to test specific hypotheses about how fish move and interact.
What's the difference between a robot that just repeats the same motion and one that adapts?
A fixed robot is like a wind-up toy—it does the same thing every time. An adaptive robot senses what's happening around it and changes its behavior accordingly, like a real fish does. If you want to understand how fish decide to follow each other in a school, you need a robot that can actually respond to neighbors, not just swim in circles.
Where does this lead? What's the next step?
Combining everything—robots that can adapt, computer simulations that model the physics and neurobiology, and new technologies that let us measure things in living fish we've never been able to measure before. The goal is to understand the whole chain: how muscles contract, how that creates motion, how motion is sensed, how sensation drives behavior. That chain operates at multiple scales, and no single tool can capture it all.