Hundreds of tiny robots can work together to grow bio-inspired shapes in a fully self-organised way.
Human designed technology usually requires an external builder. In contrast, biological systems can dynamically create their own shapes, like the shape of your hand, the dynamic displays of a flock of birds or the stripes of a zebra. The process of pattern and shape formation in nature is called morphogenesis, and it occurs in an emergent and self-organized way. “These complex behaviours are done just by having simple agents follow simple rules looking at their local neighbourhood,” explains Sabine Hauert from the Bristol Robotics Lab.
As part of an international project called Swarm-Organ (see references), Hauert and colleagues asked the question: can groups of robots be programmed like cells to grow shapes in a fully self-organised way, relying only on robot-to-robot interactions and without a map to follow?
For this study, recently published in Science Robotics, the team took inspiration from the work of computer scientist Alan Turing, who described the chemical basis for how patterns develop in nature. The process relies on a balance between the movement of chemicals by diffusion and reactions between these chemicals. As a result, patterns like spots and stripes are spontaneously created, which are called Turing patterns.
This process of morphogenesis has been translated into robot swarms formed of 300 individual, coin-sized robots. Each robot stores two virtual chemicals, one is named the activator and the other the inhibitor. Like in cells, these two virtual chemicals react with each other in a pre-programmed way. The robots also transmit these chemicals to their neighbours using infrared messaging. LED lights on each robot light up in purple, blue and then green as the activator concentration increases. As a result of the virtual chemical communication between the robots, green spots develop like Turing patterns.
The second component of the shape formation process is when the robots move and rearrange themselves. If a robot finds itself at the edge of the pack it moves along the edge and stops when it reaches a green soot. This means that from the initial tightly-packed disc the robot swarm grows protrusions, which start to reach out from the centre.
“In our shape formation everything is self-organised, thanks to these Turing patterns, and that wasn’t the case before. Those who were working on shape formation in large numbers of robots typically needed some form of coordinate system or they had a map of the shape that they wanted to create,” says Hauert.
There are several advantages to this self-organised approach to creating of physical structures. “If you look at a flock of birds, you can keep adding birds to the flock and it continues to fly, so those can work in huge numbers,” explains Hauert. “If a bird falls from the sky, the whole flock doesn’t fall to the ground, so they are also robust to individual failure and together they can do more than the sum of their parts.”
The robot swarms are also adaptable, meaning they respond to their environment and have self-healing properties. “We tried chopping off some of the protrusions and the robots either regrow the protrusions or they reassign robots that are in the surroundings,” describes Hauert.
Due to their scalability, robustness and adaptability, these robot swarms have a variety of potential future applications. For example, buildings could be designed that construct themselves and adjust their structure to the environment. Robot swarms could also be deployed to efficiently cover large areas in search and rescue situations or for environmental monitoring. But, this current study is a proof of concept and further work is needed to understand how to create functional shapes in a safe manner.