The world has changed at breakneck speed in the last 30 years. At the end of the 1980s there was no Internet or e-mail and barely any cellphones, compared to today’s hyper-connected society; a society that builds on the massive roll-out of (wired and wireless) telecom networks in the past few decades. What is interesting, though, is that this (r)evolution mainly took place between 1990 and 2005, and that there have been no large disruptions in the sector since. Michael Peeters, director of imec's connectivity program, shares with us his vision on the telecom networks of 2035 and thinks we are on the brink of a new technological revolution.
Thanks to the fixed and mobile telecom networks which seem to surround us everywhere and anytime, we have access to a wide range of communication services. We video chat with friends, colleagues or family members on the other side of the world, and we remotely control (semi-)autonomous robots (from vacuum cleaners to industrial robots).
But how should those networks evolve to deal with the needs of 2035? Let us distinguish between networks to support: (1) communication between/with people; (2) industrial applications; and (3) artificial intelligence. After that, we will briefly consider what networks might look like after 2035.
A ‘human bandwidth’ of 100 Mbps
For many people, broadband connectivity has become a given – just like electricity. Yet people who often travel – or who are often on the road – will recognize that having a stable broadband connection should not be taken for granted.
Hence, our first expectation for 2035 is that we will finally manage to build a large, stable network to support communication between people as well as human-computer interaction; a network that is no longer about the gigabit speeds that telecom operators like to boast about today, but one that is in line with our ‘human bandwidth’.
After all, whether we are talking about communication between people or human-computer interaction, both are limited to what our senses can process. Research has shown that this human bandwidth equals approximately 100 megabit per second (Mbps).
In other words: a network that can provide a continuous, guaranteed bandwidth of 100 Mbps for every user is more than sufficient to saturate our senses. And such a network should be a reality by 2035.
Of course, the question is whether we will still have the same physiological (sensory) restrictions in 2035. Maybe we will find a way round these too? Think of the neuroprobes that are being used to detect neurological diseases. In the future, we might be able to harness this technology to create direct input and output channels in our brains, meaning that wireless communication would be possible with neuroprobes in our head.
But: just because we have a direct input and output channel in our brain, and can, at least in part, circumvent our senses, this does not mean that our minds will automatically have the absorption capacity to deal with extra stimuli. So even in this scenario, there does not seem to be a need for a bandwidth of over 100 Mbps (either tomorrow, or by 2035).
And for those who are not keen on implants, smart contact lenses and augmented reality glasses will surely be available to provide extra inputs as well. Yet, paradoxically, these glasses will require more than 100 Mbps! Why? Because our senses will require the ‘right’ 100 Mbps. Think of the following analogy: people often say that a picture equals a thousand words; yet, finding the exact 1,000 words that describe a photo perfectly is much more difficult… And the same goes for our senses: the smart glasses in 2035 will have to transmit far more 'words' to provide us with the data we actually need...
Yet, whatever scenario you think is most likely to happen, there will not be any room for connections that drop out every so often – as is still the case today. Hence, we expect the telecom industry to continue to invest in (small cells) networks that can effectively provide a homogeneous – and guaranteed – speed of 100 Mbps for everyone; networks that put human bandwidth center stage, rather than today’s (peak) rates.
A flexible network for Industry 4.0 applications
In industrial settings we will face the demands of ‘Industry 4.0’ – characterized by new production methods, new products and new business models thanks to extensive digitalization.
The success of the Industry 4.0 concept hinges on the degree to which we can flexibly meet new demands and trends. As a result, a cabled environment that is difficult to reconfigure runs the risk of becoming obsolete. Yet, at the same time, wireless technologies such as Wi-Fi are often not (yet) reliable enough in production facilities with lots of reflection…
What we will need in an industrial context is a stable and reliable network that guarantees a wireless bandwidth of about 10 Mbps to and from every machine, and a limited latency.
So, in industrial environments as well, we must move towards networks that provide a guaranteed, homogenous coverage – instead of deploying unstable networks that can deliver high peak rates. This will be especially the case if one expects that human co-workers equipped with augmented reality glasses will be walking around in tomorrow’s factories as well. Technologies such as 5G, massive MIMO and distributed massive MIMO could play an important role in making this future vision a reality.
Artificial intelligence: focus on energy efficiency
The general expectation is that, by 2035, artificial intelligence (AI) will make our lives considerably easier and will help us make better decisions in all aspects of our lives. But how can we ensure that this vision becomes reality?
The secret sauce might be to ‘distribute’ the intelligence. If we want to introduce AI in an effective way, putting all the intelligence in large, central datacenters is not an option; certainly not in scenarios – such as in the case of self-driving cars – when real-time decisions must be taken, or quick anticipation is necessary. On the other hand, you cannot put the intelligence in the devices themselves (sensors, etc.), because of their limited battery and memory capacities. So, the intelligence will have to be placed in what is called the ‘edge’ of the network.
But even if putting the intelligence in the edge of the network ensures that no excessive data streams travel the network (which would require a great deal of energy and bandwidth), we still need to improve the energy efficiency of our networks by a factor of 10 to 100, to make sustainable AI feasible.
We can do so by optimizing the design of circuits and technology, but also by developing better (data) compression technology.
Will people and their habitats become an integral part of tomorrow’s telecom networks?
What the three networks that we have just described will have in common in 2035, is that they will no longer be assessed using indicators such as maximum capacity and peak rate. More important considerations will be reliability, energy efficiency and guaranteed bandwidth.
The one exception (though probably only after 2035) will be the networks that have to support truly immersive applications, such as holographic displays and virtual reality applications. These applications will require a network consisting of very small cells that can deliver dozens of gigabits per second (Gbps) at distances of just a few meters, in an extremely energy-efficient way. Think of ATTO technology.