One of hardest things about creating vast networks of devices is getting them all to communicate effectively. It turns out that fly brains may hold the solution.
Fruit flies have a highly efficient nervous system. To process the data coming in from thousands of tiny sensory hairs that provide their view of the outside world, these flies have developed a uniquely efficient way of sorting out which nerve cells need to communicate where.
Current models for computer networks are relatively fast, but require all the parts of the network to know how they’re connected to each other. The fruit fly’s method, on the other hand, doesn’t.
The way things work currently with computers, is that each cell decides if it’s a leader or not based on how many connections it has, and then all other nodes connected to it take themselves out of the running to be a leader. This continues with fewer and fewer connections required to become a leader, until everything is joined.
The fruit fly system, on the other hand, doesn’t know how many connections it has. Instead of selecting based on number of links, it selects based on time. If a cell self-selects, it sends out a signal preventing other cells around it from becoming a leader. This continues for about three hours, by which point all the cells are either leaders, or connected to ones.
While slightly slower, this method allows for an incredibly efficient network to be formed, especially when nodes aren’t distributed evenly, and may not be in communication range — like a network of underwater sensors, or a shifting wireless network. It also requires fewer messages between cells, putting less stress on the network. By analyzing the neural networks of the fruit flies, the researchers put together an algorithm to mimic it.
All this researching is being done in an effort to reach what’s known as the maximal independent set (MIS), where every node in a network is connected to a leader, but none of the leaders are connected to one another. Whether it’s millions of computers, millions of sensors, or millions of cells, the end goal is the same — create the most efficient network possible, which can function without end nodes “receiving all inputs or observing all outputs.”
Read the full scientific article published in Science
