Driving AI: Connected Cars' Highway to the Future

Facebook Twitter Email

for CNN

Updated 04:02 CET October 18, 2018

In the past decade, the automotive industry has made huge strides in connecting cars to the outside world. Developments like live traffic information, music streaming, remote starting, vehicle tracking, and navigation services have all become commonplace. But the biggest changes—and challenges—for the connected car lie ahead.

By 2023, worldwide sales of connected cars will reach 72.5 million units, up from 24 million units in 2015, according to IHS Markit, an industry consultancy. That means, in just over eight years, more than two thirds of passenger vehicles sold will be exchanging data with external sources.

The connected vehicle is nothing without communication technologies. By leveraging that connection, information can flow within the vehicle, between vehicles, between the vehicle and the road it’s traveling on and between the vehicle and the cloud—covering a range of scenarios from changing traffic lights to sudden changes in driving conditions.

This connected vehicle future, where cars can interact with the automobiles and infrastructure around them, will be a major shift says Raj Rajkumar, George Westinghouse Professor in the Department of Electrical and Computer Engineering at Carnegie Mellon University. "The most exciting piece would be when vehicles can communicate with traffic lights...and with other surrounding vehicles."

Digital highways

Some of this is already happening. Using ad-hoc mesh Wi-Fi or technologies like Cellular Vehicle-to-Everything (C-V2X), cars can communicate their intentions with one another, better avoid collisions, and learn when traffic lights are going to change. Government vehicles and roads in Colorado, for example, are already being installed with such technology. The Colorado Department of Transportation plans to equip 2,500 of its vehicles with C-V2X and DSRC connectivity by the end of the year. Within 10 years, it's hoped that up to 4 million vehicles in Colorado will be "talking" to each other and to the roadway infrastructure.

In China, the wide coverage of cellular networks and technical advances has also seen C-V2X become widely adopted by the government and it is now a key enabler of the technology. Nine provinces have embarked upon the digitization of their highways, and China aims to have 90% of the nation’s highways connected using C-V2X by 2020.

Huawei is also working with Chinese and European cities—including Barcelona in Spain, Hanover in Germany and London in England—to implement C-V2X traffic solutions globally to facilitate autonomous driving.

But this is just a start. Once connected cars become a reality, cars, pedestrians and infrastructure will be better connected, roads will become safer, engines will be more efficient, and drivers and passengers will have a better experience. As these interactions get smarter, so will AI play a greater role. Carmakers will be able monitor their vehicles in a real-world environment, tweaking the driving experience and coupling it with machine learning.

Driving interaction

AI will improve the interaction between person and vehicle, understanding and anticipating the driver’s needs and movements. Complex, data-heavy scenarios involving multiple vehicles, roads and other elements can be assessed using AI. This new branch of the Internet of Things, or IoT, will spawn business opportunities we haven’t dreamed of. Connected vehicles, says Rajkumar, may quickly become the IoT's biggest revenue generator.

The hardware and services alone that bring this functionality to cars is expected to reach $156 billion by 2020, according to a 2017 report from consultancy BCG. The number of active fleet management systems—a measure of the demand for connected commercial vehicles—is also growing, with more than 10.6 million systems covering light and heavy commercial vehicles expected in Europe by 2020, up from just 2 million in 2010.

This means a lot of data passing to and from the car. Some will be through on-the-fly connections made between vehicles, pedestrians and infrastructure as they pass. Others will be over cellular networks.

The network demands are expected to be massive: a connected car, according to Dekra, a company focusing on the safety of human interaction with technology, will generate 25 gigabytes of data per hour. , This will rise, according to chip maker Intel, to about 4 terabytes per day as cars become autonomous. That's the equivalent data traffic of almost 3,000 people.

"The network is the essential foundation for the communication between vehicles, infrastructure, and humans," says Alastair MacLeod, CEO of Teralytics, which uses mobile phone data to optimise transportation. "Not only will the network need to support billions of newly connected devices, cars and infrastructure, it will also need to be fast and reliable."

While some of this can be handled by existing LTE networks, much of the promise of connected cars will only be realized when 5G is rolled out. This could be soon. The 5G standard was completed in mid-2018 and some networks and devices are set to be rolled out by the end of the year. 5G will reduce latency and improve reliability—key for moving vehicles to communicate with a range of entities around them.

"5G enables cars to transfer great amounts of data in near real-time," explains Juri Deuter, IBM Industry Consultant and a member of its Global Automotive Center of Competence. "Also, it allows for direct peer-to-peer communication. Countries that will adopt the 5G standard quickly will have a great advantage in the whole mobility space."

The final destination

Of course, the holy grail is for cars themselves to become autonomous vehicles, or AVs. The industry divides the advancement of automotive autonomy into five levels. At the first two levels, AI assists cars, but humans are the ultimate authority. Think antilock brakes and adaptive cruise control, where humans are guided by automation, but must be prepared to intervene. The industry reached these first two levels comfortably and is now nearing level 3, says Tom Koulopoulous, a futurist and founder of the Delphi Group. "We are approaching what is one of the most critical inflection points," he says.

This level is where the autonomous vehicle can, where necessary, seamlessly hand off control to the human. "The challenge with Level 3 is that we hold the AV to a much higher standard than we do human drivers," says Koulopoulous. "For that we need better AV, which means we need better AI. And in the case of AVs, AI is still somewhat embryonic."

Driving the changes required to advance AI to this level is a focus of machine learning and its subset, deep learning, according to Huawei. "Advances in AI, especially deep learning, have propelled the automotive industry towards autonomous driving, giving new impetus to the traditional industry," said Huawei. Once these advances are achieved, Huawei believes the industry will reach levels 4 and 5, where responsibility for the car rests with the computer, and it connects with its surrounding environment seamlessly.

Even if that holy grail is some ways off, the car's dependency on data, AI and connectivity has already been established, says IBM's Deuter. "With the adoption of IoT and 5G, more data will be available and actionable. Most data loses its value within seconds. Therefore, if you are able to feed the AI application with more accurate and precise data faster, the application will improve and more use cases are possible."

HUAWEI CONNECT 2018—"Activate Intelligence"—was held at the Shanghai World Expo Exhibition and Convention Center and Expo Center from October 10 to 12.

This year's HUAWEI CONNECT conference was designed to help all businesses and organizations step over the threshold and stake their claim in the intelligent world, with by the best minds in the industry—including global ICT leaders, industry experts, and ecosystem partners—to chart the way forward and explore new opportunities.

Click here for information on HUAWEI CONNECT 2018