Robotic Fish Fins: Faster, Smarter Underwater Robots

A robotic fish hitting 1.66 body lengths per second doesn’t sound like much until you translate it into what it enables: quick, controlled movement in tight underwater spaces—exactly where bulky propellers and rigid thrusters struggle. In a recent IEEE Robotics and Automation Letters study (published September 4), a team at Zhejiang University demonstrated a flexible electromagnetic fin that propelled a small underwater robot to 405 mm/s and enabled turns with a radius of 0.86 body lengths.

This matters for a practical reason: underwater work is getting busier and more expensive. Coastal infrastructure is expanding, offshore energy assets need inspection, and environmental monitoring is moving from “occasional surveys” to “near-continuous sensing.” In our Artificial Intelligence & Robotics: Transforming Industries Worldwide series, we’ve looked at robots in factories and hospitals; underwater robotics is the same story in a harsher setting—limited visibility, high drag, unreliable comms, and tight energy budgets.

The real headline isn’t “robotic fish are cool” (they are). It’s that this fin design hints at a new path for AI-powered underwater robots: compact, agile bodies that can safely operate around reefs, pipes, pilings, and marine life while using intelligence to stretch every watt-hour.

Why underwater robots need fins (not just propellers)

Underwater mobility is a tradeoff between thrust, efficiency, maneuverability, noise, and safety. Propellers deliver strong thrust, but they bring problems that show up immediately in real missions:

• Entanglement risk from seaweed, fishing line, and debris
• Turbulence and noise that can disturb wildlife and degrade sonar sensing
• Poor performance near delicate structures (reefs, aquaculture nets, archaeological sites)
• Limited agility when you need tight turns or quick braking

Fish don’t “solve” these issues with a single trick; they solve them with compliant bodies, flexible tails, and continuous feedback control. Biomimetic underwater robots aim to steal the useful parts of that solution—especially the ability to generate thrust and steering using oscillating fins.

Here’s what most companies get wrong: they treat biomimicry as industrial design. It’s not. It’s a control-and-power problem. A fin that looks like a fish tail but can’t be controlled precisely (or drains the battery) won’t make it past demos.

The flexible electromagnetic fin: what’s new and why it works

The Zhejiang University team targeted a specific gap in aquatic robotics: actuators are often either strong but rigid (motor-driven mechanisms) or soft but weak (many soft actuators). Their fin is an attempt to get muscle-like behavior: compact, powerful, and flexible.

The actuator idea in plain terms

The fin uses:

• Two small electromagnetic coils
• Spherical magnets
• An elastic joint that supports low-friction oscillation

When alternating current flows through the coils, it produces an oscillating magnetic field. That field drives the magnets and makes the fin flap side-to-side, similar to a fish tail. When excitation stops, the fin returns to a neutral resting position.

Two performance numbers from the pool tests make it concrete:

• Peak thrust: 0.493 newtons
• Fin mass: 17 grams

That thrust-to-weight profile is a big reason this approach is interesting: it suggests you can build a small swimmer that still has enough authority to accelerate, turn, and stabilize itself—especially if you add multiple fins.

The underappreciated innovation: a predictive model

One detail in the RSS summary is easy to gloss over but should make robotics folks pay attention: the researchers built a mathematical model linking electrical input to hydrodynamic thrust output.

In soft robotics, predictability is often the weak link. Materials flex, joints creep, water flow changes, and suddenly your controller needs constant retuning. A model that lets you predict thrust from input current is valuable because it enables:

• Model-based control (more stable than purely reactive control)
• Faster design iteration (simulate before you build)
• Better energy management (know what “extra thrust” will cost)

In other words, it’s not just a fin—it’s the start of a controllable propulsion “module.”

Where AI fits: making an energy-hungry fin mission-ready

The researchers are candid about the current limitation: energy consumption. Electromagnetic coils can draw substantial current, leading to short swim times.

This is where AI and robotics integration stops being a buzzword and becomes the difference between a prototype and a product.

AI control strategies that directly reduce power draw

You don’t fix energy usage only with better batteries. You fix it by avoiding unnecessary actuation.

Practical AI-driven approaches include:

1. Event-based actuation

   • The fin doesn’t need continuous high-amplitude flapping. Use bursts to accelerate or correct heading, then coast.

2. Adaptive gait selection

   • Different oscillation frequencies produce different swimming behaviors. An onboard policy can pick “efficient cruise” vs “high-agility turn” modes based on the mission phase.

3. Model predictive control (MPC)

   • If thrust can be predicted from current, MPC can optimize a short horizon: “achieve trajectory with minimum energy,” not “track trajectory at any cost.”

4. Learning-based trim and calibration

   • Small variations in fin stiffness, magnet placement, or water conditions can be compensated with lightweight online learning so the robot stays efficient without manual retuning.

Hardware-side efficiency ideas that pair well with AI

The team mentioned directions such as coil geometry optimization and energy recovery circuits. Those aren’t just electronics tweaks; they unlock better control options:

• Optimized coils reduce resistive losses, meaning the controller can use finer control without paying a huge heat penalty.
• Energy recovery can make oscillation less wasteful—important for repetitive tail beats.
• Smart control that avoids continuous excitation is easier when the actuator and drivetrain support quick changes and predictable behavior.

A stance I’ll take: underwater robotics will increasingly look like “AI for power.” Navigation and perception matter, but the most immediate constraint underwater is still endurance.

Real-world applications: from coral-safe inspection to smart infrastructure

A robotic fish fin isn’t a novelty if it can do work where other robots fail—or do it more cheaply and safely.

Ecological monitoring without disturbing the environment

Monitoring reefs, seagrass beds, and fish populations often involves divers (costly and risky) or noisy vehicles that can disturb habitats.

A fin-propelled robot can help because it can be:

• Quieter and less turbulence-heavy than many propeller systems
• More maneuverable near fragile structures
• Potentially safer around wildlife

Pair it with onboard AI (species detection, habitat classification, anomaly spotting) and you get a mobile sensor platform that can do repeated surveys with consistent methodology—critical for environmental reporting and conservation planning.

Inspection of underwater infrastructure

Ports, bridges, offshore wind, aquaculture farms, and pipelines all need inspection. The constraints are familiar:

• Tight clearances
• Fouling and debris
• Currents and low visibility
• High cost of human-in-the-loop operations

A multi-fin system that can hold position, turn tightly, and approach surfaces gently is well-suited for:

• Close-range visual inspection
• Contact-light thickness measurements (with the right end-effector)
• Following edges and contours around pilings and joints

Search, mapping, and “boring” data collection (the profitable part)

The most scalable use case is often simple: collect consistent data over time.

A fleet of small autonomous underwater robots (AUVs) equipped with fin propulsion could:

• Map sediment changes near coastlines
• Monitor water quality near outfalls and aquaculture sites
• Inspect stormwater and underwater culverts in smart city infrastructure

This is where robotics becomes a business system: autonomy reduces operating cost, and fin-based mobility expands where the robot can safely go.

What it will take to scale: multi-fin coordination, reliability, and operations

The researchers expect the design to scale to multi-fin systems, and that’s the right next step. Real fish use coordinated fins for stability, braking, and precise turns.

Multi-fin coordination is a control problem first

With multiple fins, you can distribute effort:

• One fin handles thrust
• Another handles yaw control
• Others stabilize roll/pitch

But coordination raises practical engineering questions:

• How do you avoid control coupling (one fin’s action destabilizing another)?
• What sensors provide reliable feedback underwater (IMU, pressure, sonar, vision)?
• Can the controller degrade gracefully when a fin underperforms?

A strong input-to-thrust model helps, but robust autonomy also needs fault detection and behavior switching.

Reliability and maintainability are non-negotiable

If you’re deploying in saltwater, you’re fighting corrosion, seals, biofouling, and connector failures.

A mission-ready version needs:

• Waterproof coil and magnet packaging that survives pressure cycles
• Modular fin units that can be swapped on a dock
• Thermal management (coils + confined housings can trap heat)
• A clear maintenance schedule based on measured wear, not guesswork

In industry, these basics often matter more than top speed.

A practical evaluation checklist for buyers and builders

If you’re assessing biomimetic underwater robots (or planning a pilot), I’ve found these questions sort the demos from the deployables:

• Endurance: How many minutes at cruise, and how does it change with turning?
• Control authority: Can it hold heading in mild currents?
• Minimum turning radius: Can it navigate tight infrastructure geometries?
• Sensor stability: Does fin motion ruin camera or sonar data?
• Operate-and-recover workflow: How long does it take a crew to deploy, retrieve, recharge, and download data?

What this robotic fish tells us about AI and robotics in 2026

The fin itself is impressive: 405 mm/s speed, 0.86 body-length turning radius, and 0.493 N peak thrust from a lightweight assembly. But the broader signal is more important for leaders tracking robotics trends.

Biomimetic hardware is moving from “looks like nature” to “behaves predictably enough to control.” Once behavior is predictable, AI can optimize it. Once AI optimizes it, you get endurance, repeatability, and safer interaction with the environment—exactly the criteria that move underwater robotics into mainstream industrial adoption.

If your organization cares about marine operations—environmental monitoring, coastal infrastructure, offshore energy, aquaculture—this is the moment to start small pilots and build internal competence. Underwater autonomy compounds: every survey run improves the models, the maps, and the mission planning.

Where do you want your underwater data to come from in two years: occasional, expensive field trips—or persistent robotic sensing that gets smarter every month?