In 2004, a few years after the first 3G service was launched, my wireless professor Dr. Arslan told the class that 3G was an amazing technology that needed a killer app for a phone that could use high data rates. Phones of that time couldn’t run a web browser. Several years later, operating systems geared toward “web apps” became the killer app that put higher data rates to good use. 
A year ago a colleague showed me his 4G phone doing 20Mbps on the downstream. I found this amazing. 4G potentially supports higher data rates.
Standards bodies are now working out specifications for 5G. They aim to finalized the specs by 2018 and do initial deployments in 2020.
Apps demanding more bandwidth drove 3G and 4G, but the drivers for 5G will be more diverse. The biggest driver may be dealing with the explosion of data usage. This month’s IEEE Communications Magazine focuses on the question of 5G.
An article from Alcatel-Lucent Bell Labs proposes five technology directions that may define 5G:
1. Device-Centered Architectures - 5G may allow mobile units to connect to a different base station for upstream and downstream data. There is also the whole question of handling heterogeneous networks, i.e. bases that cover large areas alongside tiny pico- and femto-cells, intelligently. 5G may support multihop, allowing a user inside a building to connect to a base station through someone else’s mobile. This feature has been discussed for the past ten years; I’m interested to see if it will ever be realized.
2. Millimeter Wave (i.e. 3 to 300 GHz) - These frequencies sound insanely high. The path costs are higher for line-of-sight links, and obstructions block the signals much worse than on lower bands. I remember reading about 1GHz in a communications magazine in the 80s. It said at these frequencies radio waves “take on some of the properties of light” and can be reflected by a piece of metal. Almost 30 years later, I still remember it. I thought this would make frequencies over 1GHz useless for terrestrial communication, which turned out to be completely false. I suspect mobiles operating in the 30GHz range will appear eventually, although I’m skeptical it could happen by 2020.
3. Massive MIMO - MIMO is the most amazing thing I have seen in wireless technology, in that it allows multiple streams of data in the same place at the same time on the same frequency. The IEEE article mentions using four-stream MIMO, but I am confident this will not happen. For MIMO to work, each stream needs an antenna separated from the others. Two antennas with orthogonal polarizations could allow for two streams. To go beyond that, the antennas need to be separated by more than half a wavelength, and it really helps if the ratio of link path to antenna separation is not too high.
What I do see happening with 5G is beamforming. Base stations with multiple, well-separated antennas can use beamforming to act as a highly directional antenna. This seems very likely to be part of the standard because it does not require the mobile to have separated antennas.
4. Device-to-Device (D2D) and Local Caching - Engineers have long recognized the inefficiency of people sitting nearby each other sharing data through a distant base. It seems likely that some provision for transmitting data directly to other users will be part of the standard. 
Those of us who have intermittent data coverage know which of our apps are smart enough to cache data when there is no coverage and sync up once connectivity is resorted. I don’t know if this will be part of the 5G standard, but I expect to see more apps supporting it.
5. Support for Machine-to-Machine (M2M) - Most data service plans focus on throughput, but latency is a bigger issue for many applications, especially M2M. Industrial protocols often use short packets but need a fast reply. Data rate isn’t important because the packets are so small. The tricky part is quickly giving channel access to a nodes with data to pass.
A surprising number of industrial users (e.g. factories, pipelines, sewer and water systems) use Wi-Fi or frequency hoppers in the same bands as Wi-Fi to link machines to a human-machine interface. With good antennas, these links are reliable over several miles. The need to go beyond this range and issues associated with using the ISM bands for important data motivate M2M users to use the mobile data network. More M2M traffic will move to the mobile data network 5G decreases latency.
Conclusion
Although it’s not highlighted in the IEEE articles, most proposed 5G features are geared toward handling the explosion of mobile users. Users outside densely populated urban areas will see little benefit from 5G features.
I expect rural areas will not adopt 5G for a long time. There the predominant issue will be filling the role of the publicly switched telephone network, which is a political issue more than a technical one.
