Machine learning to accelerate green internet development

Machine learning and artificial intelligence are becoming increasingly important tools in the development of the energy efficient internet and data transmission of the future. DTU is strengthening the area with the appointment of Group Leader Darko Zibar as Professor at DTU Fotonik.

It is a well-established fact that machine learning and artificial intelligence can find connections in large data sets—e.g. in the health sector—but it is relatively new that researchers are beginning to see how the methods can be applied in developing energy-saving solutions in such fundamental internet technologies as optical communication and intelligent optical systems.

“Next generations of optical communication systems will be so complex that machine learning will be a relatively quick way to find solutions that can transport huge amounts of data in the most energy efficient way”

Professor Darko Zibar, DTU Fotonik

The internet currently consumes about nine per cent of the world’s total electricity consumption—and about two per cent of man-made CO2 emissions. Already within the next decade, the need for data capacity and bandwidth will increase dramatically as a result of all the things society is constantly connecting to the internet. There is therefore a need to find ways of accelerating the development of energy-efficient technology that will ensure a green internet in the future—and newly appointed professor Darko Zibar is in no doubt about what can drive development forward.

“Machine learning and other intelligent systems can play a key role, and we therefore need to research this field if we are to achieve the UN’s SDGs on sustainable development and the reduction of global CO2 emissions. The next generations of optical communication systems will be so complex that machine learning will be a relatively quick way to find solutions that can transport huge amounts of data in the most energy efficient way,” says Darko Zibar.

Intelligent systems essential for development

The advantage of machine learning is that the computer is able to analyse huge amounts of data and find algorithms and connections without being pre-programmed.

“We can use machine learning to identify models that can describe the relationship between transmitter and receiver when we develop lasers, frequency combs, networks, etc. to transport large amounts of data. Indeed, there are many optical system factors that need to be assessed in the hunt for the most energy-efficient solution—e.g. bandwidth, channel power, frequency noise, traffic route, and much more,” says Darko Zibar.

He explains that the increasing focus on quantum technology to improve internet security makes the task of designing optical communication solutions even more challenging because it requires coexistence and control of classic channels and quantum channels in the same optical network. This creates the need for the development of intelligent optical receivers that can distinguish between classical signals and quantum signals—so here too it is necessary to use machine learning and artificial intelligence in the development process.

“There will also be potential in using artificial intelligence in optical measurement systems—e.g. current optical instruments used to analyse the quality of the signal can’t distinguish between different disturbances and determine whether they originate from the transmission channel or the components themselves. If signal analysers can learn to distinguish between the different disturbances, we will be able to design more efficient transmitters and receivers from the signal processing algorithms.”

Developing efficient optical communication systems and networks is demanding, and many initiatives using different technology approaches are needed. That said, the need is urgent, and Darko Zibar and his group are therefore actively researching areas where they believe machine learning and artificial intelligence will have an impact on next-generation optical communication systems.

/Public Release. This material comes from the originating organization and may be of a point-in-time nature, edited for clarity, style and length. View in full here.