Machine learning can help devices maintain contact in emergencies.
Electronic communication forms such as text, image sharing and social media rely on robust and reliable networks. But those networks can break down during natural disasters, power grid failures or cyberattacks—making it much more difficult for people to communicate with each other in potentially life-threatening situations.
Researchers from the Department of Energy’s Oak Ridge National Laboratory (ORNL) are working to fix that problem by teaching devices to communicate with each other even when their digital networks are unavailable. They are programming machines to create impromptu language to speak to each other through alternative transmission options such as acoustics, optic or radio frequencies.
“We’re training and programming devices to figure out the best way to communicate for themselves,” said Adam Anderson, who leads a computing, networking and cybersecurity team at ORNL.
A key obstacle the team had to overcome was teaching the devices how to learn to communicate with each other independently without any outside help.
“We felt it was ‘cheating’ if the devices learned anything from an outside, or unrealistic, source; they need to converge all on their own,” Anderson said.
To demonstrate how devices could learn to speak to each other, the researchers set up two laptops facing each other, connected to separate speakers and microphones. These laptops were programmed with language that uses sounds—such as the consonants, vowels and syllables of human speech—rather than the digital bits computers usually use.
One computer was instructed to “speak” the numbers one to five to the other; the receiving laptop then outputted “won, too, three, for, five” on its screen. As the computers “spoke” to each other they adjusted and correlated their speech until the receiving computer produced the numbers correctly.
Teaching machines how to transmit and receive messages on their own can make emergency communications more robust. For example, walkie-talkies, which are limited to a specific range of available frequencies, could be programmed and trained through machine learning to seek other ways to keep first responders connected if those frequencies are unavailable.
The devices could “begin sharing data of any kind over just about any physical medium,” according to Anderson.
“They will initially go through a trial-and-error volley of information, developing their own patterns as they go until they reach the most efficient and accurate mode of message transmission,” he said.
The ORNL team hopes to use their findings to replace classic digital modems in computers, which would make communications networks more resilient to disruptions and malicious attacks—and would help people communicate better in times of crisis.
To learn more about machine learning, check out 5 Machine Learning Trends Will Train the IoT Market.