A new computer technology based on the nervous system of fruit flies could lead to more efficient wireless networks.
The tiny, hairlike tentacles that line the fruit fly’s nervous system act as sensing devices. This sensing system is organized in such a way that just a few “leader” hairs communicate with every other hair cell, giving the insect a quick big picture of the world without having to know where each individual hair is located.
Taking the lead from this insect system, scientists have developed an algorithm that could be used to improve the communication between swarms of robots and other wireless networks. For example, lots of tiny sensors could be dropped from an airplane or helicopter to monitor environmental hazards such as volcanic eruptions.
“There is no initial knowledge of how they are scattered around and so they need to initially establish some backbone for a network, which is what our algorithm does,” said Ziv Bar-Joseph, associate professor of machine learning and computational biology at Carnegie Mellon University.
Bar-Joseph said that the computer model is ready to use and just needs to be plugged into a particular application.
Distributed computer networks
While desktop computers have one central processing unit that listens and responds to the rest of the processors in the computer to make decisions, distributed networks have a bunch of so-called leaders.
“Distributed computing is an area within computer science that deals with how lots (hundreds, thousands or even more) of computers can work together to achieve a common goal,” Bar-Joseph said.
One example of this type of system is Google, where thousands of servers interact with one another to answer users’ queries. To make these Internet searches quick, Google has selected a small group of leaders from the large number of available servers. (Every server connects to one of these leaders and these leaders communicate with one another to retrieve the best information for your query.)
“So far, for computer networks this selection of local leaders relied heavily on knowledge of your neighborhood (how many other devices are within a broadcast range from you),” Bar-Joseph said. “Such knowledge is often hard to obtain since in many applications computer sensors are dropped off in a specific location with no initial knowledge of who their neighbors are,” said he added.
Uncovering the fly’s secret
The fruit fly, however, can do this selection process without prior knowledge of the location of its individual hairs. Also, “the fly is very efficient sending very few small messages to perform this process (as opposed to many more messages required by current computer methods),” Bar-Joseph said.
To mimic the fruit fly’s nervous system, the researchers watched the insect’s selection process under a high-magnification microscope and figured out how the fly selects its leader hair cells. Next, they modified the process to allow for more processors as is typical of a computer network.
The scheme is “ready to go” and waiting for someone to use, Bar-Joseph said.
Bar-Joseph and his colleagues detailed their findings in the Jan. 14 issue of the journal Science.
