A moth-shaped robot used an artificial circulatory system to deliver chemical building blocks through its body and grow a light sensor where UV light struck.
A robot's brain can only do as much as its body allows. Robots are increasingly sent into places nobody can fully map in advance, such as collapsed buildings, forests, the deep sea, or other planets. Yet their hardware is fixed at the factory: every sensor must be put in place before the robot leaves the lab, based on a guess of what it will encounter. The current state of the art is an adaptive algorithm in a centralised controller, but this does not fully compensate for a robot body poorly matched to the unexpected environment. A possible solution is to create new components as the situation unfolds, thereby enhancing the robot's adaptability.
Building a new component from scratch takes two things: suitable building blocks, and a way to get them to the right place at the right time. Nature solves both with circulation, as blood or hemolymph carries building blocks throughout the body, ensuring that each organ has access to new materials and can dispose of used ones when needed.
Researchers at the University of Tartu in Estonia and the Italian Institute of Technology set out to give a robot the same ability. "I'm working on giving robots a circulatory system, so they can transport molecular building blocks across the whole body and create the hardware they need, on site, where and when it's needed," says lead author Kadri-Ann Pankratov.
The team took its blueprint from moths. Like other insects, moths have an open circulatory system: a pumping vessel stirs the hemolymph, the insect's blood, which loosely follows a few channels but mostly just bathes the organs inside the exoskeleton. "That makes the architecture simpler. You can focus on the fluid content instead of optimising its flow," Pankratov explains. Insect wings are also living structures filled with hemolymph, and their nearly flat shape allows the fluid to flow in a single plane, keeping the robot's plumbing simple. As a model, the researchers used the blue underwing (Catocala fraxini), one of the largest moths in Estonia, with specimens provided by colleagues from the University of Tartu Institute of Ecology and Earth Sciences.
The robot's body was 3D-printed from a transparent plastic (PETG) and operates on two scales. Millimetre-wide "veins" carry fluid across wings of 86 square centimetres each, while the gaps between the printed lines form a dense network of microchannels from which the liquid soaks about 10 micrometres into the plastic itself. A single pump with open-loop control fills the entire body with about 3 millilitres of liquid.
To make a sensor, the team pumped an "artificial hemolymph" through these veins: a mixture of pyrrole, a light-activated initiator, and a cellulose-based additive. Wherever UV light then shone on the body for about a minute, the pyrrole inside the plastic turned into polypyrrole, an electrically conductive polymer. The synthesis of a new sensor was also evidenced visually, as the translucent body turned black. Elsewhere, it stayed unchanged. Using a stencil, the team could draw sensor patterns with 0.3-millimetre precision.
The new material works as a UV receptor: when UV light hits it, its electrical resistance briefly drops. Before the sensor was grown, the robot did not react to UV at all. Afterwards, a small onboard microcontroller detected the change and made the robot flap its wings, driven by a shape-memory wire, or blink an LED. The robot thus closed a sense-and-react loop with hardware it did not have when it was built. "It's not about reviving a lost capability, the way a lizard regrows its tail, but about creating a completely new, previously non-existent body part in the robot," Pankratov says.
Why grow a sensor rather than install one? "You can, of course, use a ready-made photodiode, but you'd have to know quite precisely where to put it. That can be difficult to know beforehand as robots go into more dangerous and complicated environments," Pankratov says. "Growing the receptor on site means the environment itself can pattern the robot, based on what is actually happening at that time."
One finding was unexpected. The sensor did not behave like a constant, factory-calibrated component: with each exposure it became more sensitive, as leftover pyrrole kept polymerising. "It wasn't planned, but it's very logical that it happened. The conductance kept changing in one direction, so it was clear that something beyond ordinary drift was going on," Pankratov says. The authors interpret this as a form of material memory: the sensor's structure becomes a record of what the robot has experienced. The sensors still worked after more than four months of storage.
The research suggests new pathways to give the physical embodiment, which works closely with a centralized controller, more autonomy. The fact of installing a new sensor, its position, time of installation, and even its sensitivity were all determined by interaction with a prepared body with its environment, not premediated. However, the butterfly still has one important lever to control its behavior: when and how to prepare the body for environmental encounters. The choice of a chemical to be broadcast in the circulatory system also determines the type of response upon an encounter. This parallels natural circulatory systems: for example, in a stressful situation, the insect brain can trigger the release of octopamine into the circulatory system, altering the body’s preparedness for new environmental encounters.
The study is a proof of concept with clear limits. It demonstrates a single, predetermined chemical pathway, and the synthesis is irreversible. In the experiments, the researchers decided when to fill the veins, and a stencil, rather than the natural surroundings, decided where the sensor formed. The light that created the sensor was also the signal it later detected, although the authors argue that the approach extends to other chemicals and triggers. Pankratov sees the biggest open question elsewhere: "Right now, the robot's central intelligence knew which component was coming. How do we make it so that it doesn't have to be predetermined? We don't yet have a good source of inspiration in biology for that."
The team expects the toolbox to grow. Polypyrrole alone can be tuned for corrosion protection, electromagnetic shielding, or interfacing with living tissue, and the same veins can carry many other dissolved or dispersed chemicals, possibly including reversible ones. "In ten years, I hope artificial robot blood and circulatory systems will be an obvious choice for some applications, with several material systems as the state-of-the-art setup," Pankratov says. She imagines such robots working in natural habitats.
The study "Receptogenesis in a Vascularized Robotic Embodiment" was published open access in Advanced Science 23 Oct 2026 issue: https://doi.org/10.1002/advs.76999 , recognised with a Frontispiece https://doi.org/10.1002/advs.78174
The research was supported by the Estonian Research Council grants PRG1498 and TEM-TA56.
Advanced Science
Experimental study
Not applicable
Receptogenesis in a Vascularized Robotic Embodiment
23-Oct-2026
The authors declare no conflict of interest.