A team led by the University of Oxford has developed a new class of soft robots that operate without electronics, motors, or computers - using only air pressure. The study, published today (05 Nov) in Advanced Materials , shows that these 'fluidic robots' can generate complex, rhythmic movements and even automatically synchronise their actions.
Professor Antonio Forte (Department of Engineering Science, University of Oxford, Lead of RADLab) said: "We are excited to see that brain-less machines can spontaneously generate complex behaviours, decentralising functional tasks to the peripheries and freeing up resources for more intelligent tasks."
Overcoming a key challenge in soft robotics
Soft robots (made from flexible materials) are ideal for tasks like navigating uneven terrain or handling delicate objects. A major goal in soft robotics is to encode behaviour and decision-making directly into the robot's physical structure, enabling more adaptive and responsive machines. This kind of automatic behaviour - emerging from body-environment interactions - is often difficult to replicate with traditional electronic circuits, which require complex sensing, programming and control systems.
To address this challenge, the researchers took inspiration from nature, where body parts often perform multiple roles and synchronised behaviour can emerge without central control. Their key innovation was to develop a small, modular component that uses air pressure to perform mechanical tasks - similar to how an electronic circuit uses electrical current. Depending on how it is set up, this single block can either:
- Actuate (move or deform) in response to air pressure changes - functioning like a muscle.
- Sense pressure changes or contact - similar to a touch sensor.
- Switch air flow between ON/OFF states - like a valve or a logic gate.
Similar to LEGO pieces, multiple identical units (each one a few centimetres in size) can be connected to form different robots without changing the basic hardware design. In the study, the researchers constructed tabletop robots (roughly the size of a shoebox), that could hop, shake, or crawl.
In a particular configuration, the researchers found that each individual unit can automatically combine all three roles at once, enabling it to generate rhythmic movement entirely on its own once constant pressure is applied. When several of these responsive units are linked together, their movements began to synchronize naturally, without any computer control or programming.
These behaviours were used to make a shaker robot (able to sort beads into different containers by tilting a rotating platform) and a crawler robot (which could detect the edge of a table and automatically stop, preventing a fall). In each case, the coordinated movements were achieved entirely mechanically, with no external electronic control.
Lead author Dr Mostafa Mousa (Department of Engineering Science, University of Oxford) said: "This spontaneous coordination requires no predetermined instructions but arises purely from the way the units are coupled to each other and upon their interaction with the environment."
Laying the groundwork for embodied intelligence
Crucially, the synchronised behaviour is only seen when the robots are linked together and touching the ground. The researchers used a mathematical framework called the Kuramoto model, which describes how networks of oscillators can synchronize, to explain this behaviour.
This revealed that complex, coordinated motion can emerge in the robots purely from their physical design when they are mechanically coupled through the environment. In this case, the motion of each robotic leg subtly affects the others through the shared body and ground reaction forces. This creates a feedback loop where the forces transmitted via friction, compression, and rebound link the motions of the limbs together, leading to spontaneous coordination.
Dr Mousa said: "Just as fireflies can begin flashing in unison after watching one another, the robot's air-powered limbs also fall into rhythm, but in this case through physical contact with the ground rather than visual cues. This emergent behaviour has previously been observed in nature, and this new study represents a major step forward towards programmable, self-intelligent robots."
Although the soft robots developed are currently at tabletop scale, according to the researchers the design principles are scale-independent. In the near future, the researchers aim to investigate these dynamical systems to build energy-efficient untethered locomotors. This would be one step forward towards the large-scale deployment of these robots in extreme environments where energy is scarce and adaptability is needed.
Professor Forte added: "Encoding decision-making and behaviour directly into the robot's physical structure could lead to adaptive, responsive machines that don't need software to 'think.' It is a shift from 'robots with brains' to 'robots that are their own brains.' That makes them faster, more efficient, and potentially better at interacting with unpredictable environments."
