5G & Edge Computing: Enabling Autonomous Vehicles

A car moving at 50mph travels 72 feet — or 22 meters — in one second. If something happens on the road, and the vehicle’s network sends a message to the brakes that takes 100 milliseconds to arrive, it could be too late to stop an accident. Any delay could be critical.

Now imagine a self-driving vehicle, collecting millions of data points per second and receiving several thousand more from other vehicles and the road, sending all that information to the cloud for processing and waiting for a response.

Even if speed and latency are not an issue to connect to the network’s basestation, transferring all that data to the cloud will take several milliseconds. Then it has to processed, and the resulting action sent back to the vehicle.

If that means that breaks are not activated in time, it could mean the difference between safety and a disaster.

Bandwidth limits
Today’s wireless networks have run into a problem — more people and devices are consuming more data than ever before, but data remains crammed on the same bands of the radio-frequency spectrum that mobile providers have always used. Those bands, from 300KHz to 6GHz, are used for everything: smartphones, fixed wireless phones, remote control devices, all WiFi 801.11 specs, Bluetooth and other devices.

That means less bandwidth for everyone, causing slower service and more dropped connections.

One of the key advantages of the upcoming 5G networks is the use of much higher frequency bands — “Millimeter Waves” — which broadcast at frequencies between 30 and 300 gigahertz. These frequencies, due to the higher spectrum and wider channels, offer much faster transmission rates and extremely low latency — up to a 100 times lower than current 4G networks.

However, there’s a catch.

Millimeter waves are unable to penetrate walls. Their effective range is limited to a few hundred meters, and they can be absorbed by trees, pollen and high humidity. In fact, to deliver the high bandwidth and low latency 5G promises, these waves need to be much closer to every device on the network.

Once the first 5G networks start to appear in cities, we’ll start to see small cells pop up everywhere to compensate for the lack of signal penetration. And we shouldn’t expect to see much 5G coverage outside dense populated areas.

Cruising to the edge
Having gigabit speeds and low latency, down to one millisecond, is not enough to ensure quick response, especially on self-driving vehicles. Fast on-board CPUs, plus additional processing at basestations, is a must.

This is commonly referred to as edge computing.

And the connected-car industry is embracing it. Last year, Denso, Ericsson, Intel, Nippon Telegraph and Telephone Corporation (NTT), NTT Docomo, Toyota InfoTechnology Center and Toyota, created the Automotive Edge Computing Consortium.

To understand more about how edge computing enables connected cars and its implications for 5G, I talked to two experts during this year’s Mobile World Congress expo in Barcelona: Thomas Henze, the head of mobile access at Deutsche Telekom, and Sandro Tavares, the head of cloud marketing at Nokia.

In an interview, Henze said that operators are looking for new ways to enable processing close to the connected device, something that will be critical when 5G networks start to operate. Deutsche Telekom is already installing basic processing, such as channel aggregation, on LTE networks, and will expand it to the upcoming 5G basestations and routers.

Tavares noted that 5G technology offers another view of the need for edge computing in the area of connected cars.

It is impossible to keep the low latency of the network if processing needs to be outside the connected device, Tavares said. Instead, 5G networks will offer unparalleled reliability and faster transmission speeds but the connected devices should be able to function independently of the data centers, otherwise the advantages of the new frequencies will be lost.

It all comes down to transmission and memory speed.

Moore’s Law, the rule that the number of transistors on chips doubles every two years, has been mostly accurate until now. Memory speed and performance, however, haven’t improved at that rate and, relatively to processing time, the cost of accessing memory has increased exponentially.

That is why we have caches for caches for caches — all the way down. And those caches need to be positioned within the devices collecting and processing data.

Future of driving
With onboard vehicle-to-vehicle (V2V) and vehicle-to-everything (V2X) communications using a 5G network, connected cars will be able to share information on the fly between the vehicle and the infrastructure with almost no delay and at previously unseen speeds.

When a self-driving vehicle receives information from the infrastructure about a traffic jam, or its on-board Lidar sends a warning about an obstacle, the main CPU can take immediate, life-saving action.

This concept of edge computing is not only changing the way connected vehicles are being developed, but also other application that require a quick local response.

“By 2022, as a result of digital business projects, 75% of enterprise-generated data will be created and processed outside the traditional, centralized data center or cloud, which is an increase from less than today’s 10%,” according to Gartner.

Imagine the chaos on the streets if connected and autonomous vehicles need to rely on cloud computing server farms, and what could happen if those servers went down. A hacker accessing the servers could stop or take control of the vehicles or the infrastructure connecting them. That is why the market needs edge nodes distributed on the cars and the streets, and on-board local processing to take the right decisions on the spot.

— Pablo Valerio is a technology writer and consultant working out of his home city of Barcelona, Catalonia. Follow him on Twitter @Pabl0Valerio.

Leave a comment

Your email address will not be published. Required fields are marked *