When 5G, the fifth generation of wireless communications technology, arrives in 2020, engineers expect that it will be able to handle about 1000 times more mobile data than today’s cellular systems. It will also become the backbone of the Internet of Things (IoT), linking up fixed and mobile devices—vending machines and cars alike—becoming part of a new industrial and economic revolution, some say. A new architecture, new communication technologies, and new hardware will make this transformation possible. A research team comprising Zhiguo Ding at Lancaster University and researchers at China's Southwest Jiaotong University have took stock of the recent research and future needs of 5G in a review last month in Science China [pdf]. In a Skype interview Ding spoke to Spectrum about his views on how 5G will take shape in the near future.
Spectrum: 4G is now being deployed in many countries, but is it already out of date?
Ding: Actually 4G is good for now, however if you would look at it in five or ten years, 4G will obviously not be able to meet requirements for new applications coming up in the next few years. With 5G we will increase the data rate, reduce the end-to-end latency , and improve coverage. These properties are particularly important for many applications related to IoT. One example is emerging autonomous cars and intelligent transportation, to which small latency is essential. Another example is that a lot is happening with interactive mobile games, which are really bandwidth hungry. Unfortunately the current 4G cannot support them.
So 5G will play a fundamental role in the Internet of Things?
Ding: I think the Internet of Things will be the ideal application for 5G. What currently stands in the way of the IoT are disconnected systems. For example, we have RIFD, we have short-range communications techniques, UWB, Blue Tooth, etc., and this could be a problem in the future if we talk about a bigger picture like a smart city, where a unified framework for seamless connection is required. 5G is a good opportunity to provide this unified framework.
How will 5G deal with the huge number of devices connected to the IoT? Will there be sufficient bandwidth?
Ding: The previous 1G-4G systems rely on so-called orthogonal multiple access. Take time division multiple access used by 2G as an example: We slice one second into a lot of timeslots with short duration. We then allocate one particular time slot to each user, and one user cannot access a channel allocated to others. Such orthogonal multiple access will be difficult to support for future IoT applications. We will have a lot of devices, and we would have to allocate time slots dedicated to each of them. But in the end this is a luxury we cannot afford, since the number of available time slots and bandwidth resources will be insufficient. This is why multiple orthogonal access won't work for the 5G.
Currently there is a lot of research exploring how we can develop non-orthogonal multiple access by putting a number of users into the limited bandwidth channels. Ideally non-orthogonal multiple access can strike a better trade-off between system throughput and user fairness. Of course there is interference between users, which means that some users may experience low data rates. But interestingly in IoT, there are many devices which should be served timely with low data rates. One example is wireless healthcare, where wearable devices (heart monitors, biosensors, etc) need to send patient data timely to hospital severs, but the data rates used by these devices are not likely high. By using non-orthogonal multiple access, we can squeeze in a lot of IoT users/devices with different quality of service requirements into the same time slot or frequency channels. In this sense, the concept of non-orthogonal approaches is very exciting and perfect for the Internet of Things.
Another way to illustrate the benefit to break the orthogonality of multiple access is to view non-orthogonal multiple access as the special case of cognitive radio technologies. Currently, we allocate a single bandwidth, or channel, to a user, and we are not able to reuse it again because this user is occupying this channel. With cognitive radio communications we can admit new users into this channel. If these users have good connections to the base station, we can realize a large data rate. Of course this will cause some performance degradation for the initial users, but such degradation can be insignificantly if these initial users have poor connections or a careful power control mechanism is carried out among the users.
How is 5G going to deal with the "spectrum crunch"—the available RF bands are becoming full?
Ding: To solve the spectrum crunch we need a combination of a lot of technologies. One way is to improve the efficiency for using the existing available bandwidth. In this sense, we can apply non-orthogonal multiple access, massive MIMO, cloud radio access networks, full duplexing, etc.
Another way is going to shorter, milimeter wavelengths, 60 or 90 GHz, where more spectrum bandwidth is available for telecommunications. There are some challenges here. For example, the higher the frequency, the more the attenuation by the atmosphere, and this rules out long-distance transmission. In addition, there is the shadowing problem—you need an intact line of sight between the transmitter and receiver. This problem can be solved with multiple antennas; you have backup links between the transmitter and receiver, even if one link is blocked. It is worthy pointing out that the use of millimeter-wave communications is promising for many applications in IoT since sensors might have line-of-sight connections and distances among sensors might not be large.
What are the next steps?
Ding: The timeline for the development of 5G has not been officially confirmed. It is widely expected that a formal discussion as well as standardization activities will start in the next year, and commercial deployment is expected to happen in 2020. Currently both industry and academia are working together to identify which standards and techniques should be used and which not.
