中文 English

L5 Adoption Hinges on 5G/6G

Wireless technology is key to both inventing and selling fully self-driving cars.

popularity

Truly self-driving cars don’t yet exist, and research shows many consumers are wary of them anyway. What will it take to make fully autonomous cars possible? And how can automakers convince consumers to adopt such vehicles?

Experts say the answer to both questions could lie in wireless communication networks. That’s because such networks offer a workaround to a major obstacle in autonomous vehicles — fitting sufficient compute power into a machine with significant weight, power, heat, safety and cost restrictions.

Today’s most advanced vehicles, which fall short of Level 5 autonomy, use a fourth-generation network to navigate through traffic, stream music and videos, and download updates that impact the function of the car, among other activities. Fifth- and sixth-generation networks are expected to enable these activities to be completed faster and more securely, while also enabling vehicle-to-vehicle communication and communication between the car and smart city infrastructure — features that will be critical to reaching L5.

In addition, 5G and 6G networks will help fully autonomous vehicles become mainstream. Humans will be less anxious about ceding control of a car to a computer, or so the theory goes, if the vehicle is able to quickly and clearly communicate with its passengers about what it is doing and why. Also, consumers will be more likely to embrace these vehicles if they are able to use the network to create a more custom and entertaining experience than can be found in a car with a lower level of autonomy.

Of course, no one knows what exactly it will take to reach L5 or mainstream adoption. Technical challenges, like navigating through a snowstorm with blocked sensors and social challenges like government regulation are not easily solved with wireless networks. But Thierry Kouthon, technical product manager at Rambus, says he views 5G/6G as part of a natural extension of trends that have dominated the industry for the past 10 years. According to Kouthon, those trends are advanced driver assistance systems, electric vehicles and autonomous driving.

“They all have the same thing in common,” said Kouthon. “They are trying to make our cars clever. Environmental awareness is fundamental for these new vehicles to make correct decisions. The combination of 5G/6G and vehicle-to-vehicle promises to be essential ingredients to enable the roadside infrastructure such as traffic lights, railroad crossings, and vehicles themselves to communicate with other vehicles and share a common view of traffic that will dramatically reduce fatalities. 5G/6G technologies will be essential to generate the breeding ground to create an adequate inter-vehicular network conducive to a more harmonious and safer traffic experience.”

Hanging in the balance is more than a trillion dollars in revenue, as well as a significant number of human lives. McKinsey & Co. estimates that global revenues associated with autonomous vehicles in urban areas could reach $1.6 trillion by 2030, more than double the combined 2017 revenues of Ford, General Motors, Toyota and Volkswagen. It is more difficult to predict the number of lives self-driving cars would realistically save, though the World Health Organization says 1.3 million people die each year in road deaths. McKinsey calculates that full adoption of autonomous vehicles in the U.S. alone would create a public benefit of $800 million per year in 2030 in the form of redevelopment of unnecessary parking spaces, more productive commute time for workers, reduced environmental damage, and safer roadways.

Achievements
Realizing any potential benefit to businesses or the public requires automakers to produce self-driving cars that function as intended. There are currently no L5 cars on the road, and many experts predict the technology is at least a decade away from viability. One of the major challenges is how to cram enough compute power into such a smart vehicle without compromising the function or safety of the car, or making a machine so expensive it could never possibly be available to consumers.

David Fritz, vice president of hybrid-physical and virtual systems for automotive and mil/aero at Siemens Digital Industries Software, said it appears the industry is moving toward solutions that involve relatively “stupid” cars with more compute power in the cloud. “What we’re seeing is a clustering towards having plenty of capability on vehicle, and augmenting the sensory information from what’s happening in a 5G or 6G cloud,” he said. “It could be from another vehicle, a satellite, or smart city infrastructure.”

Communication between vehicles also will be an essential function of a wireless network in this version of the future. An ambulance, for example, will need to tell the other vehicles to wait for it to pass regardless of other traffic information. Fritz said it’s not yet totally clear how this will work. “It has to be a high-powered broadcast to any car within a reasonable vicinity, and some of that broadcast has to be through buildings in an urban environment,” he said.

Keysight Technologies offered several predictions for what that experience might look like:

  • Vehicles that are connected to each other, road infrastructure, and cloud-based services;
  • Advanced connectivity with vulnerable road users, like cyclists and pedestrians;
  • Reduced need for large parking areas because cars can park themselves and don’t need space for a door to be opened, and
  • Integration with satellites and unmanned aerial vehicles.

“6G will enable significantly enhanced integration of communication systems, sensing technologies, services, and human experience, with the vehicle taking on the role of a digital experience center connected to an evolved and intelligent network,” said Malcolm Robertson, 6G strategic planning manager at Keysight. “Artificial intelligence and machine learning will play a central role in providing on-the-fly services and capabilities, synthesizing and fusing data from multiple inputs, and optimizing for safety and efficiency. New capabilities will emerge such as the ability of a vehicle to ‘see’ around corners and through obstacles, correctly identify objects and scenes in its fields of view, navigate effectively in all weather, as well as communicate seamlessly with its environment and other connected entities.”

This will require some technology shifts. “Mobile network operators are re-farming frequency bands that have been used for legacy 2G and 3G technologies to 5G,” said Bill McKinley, automotive/connected car business lead at Keysight. “This has the knock-on effect of vehicles not being able to use the circuit-switched technologies for applications such as emergency calling, which is a legal requirement in Europe. The solution for this is the adoption of next generation emergency call (NGeCall) in vehicles, which uses packet-switched networks such as 4G and 5G.”

That’s just the beginning. “In-vehicle architectures are being redesigned to allow for higher-bandwidth, lower-latency communications between the numerous processors that comprise the modern vehicle,” McKinley said. “All these changes require advanced testing and new test tools. Due to the complex nature of testing advanced applications, it can be expensive and logistically challenging to use real vehicles on test tracks or on real roads. Therefore, increasing the use of realistic simulations to cover as much test as possible before final road testing is an appealing proposition.”

Regardless of whether it is 4G, 5G, or 6G, wireless technology presents its own challenges that must be overcome. For vehicles to communicate with one another, for example, there will have to be some amount of standardization in the tools the cars use to “talk” over the network. And who will pay for smart infrastructure throughout the world that will allow autonomous vehicles to communicate with their surroundings?

Hyper-connected vehicles also raise new questions about how to keep data safe and private over the network, as well as how data security could be compromised, and in turn impact the safety of the humans in and around the car.

“There’s always the issue of doing remote firmware updates in a secure manner,” said Maarten Bron, Riscure’s managing director. “Indirectly, the prevalence and performance of 5G will result in more devices being connected, and from an attack standpoint this enlarges the so-called attack surface. Ergo, more devices to be hacked.”

Security experts say each generation of wireless technology is more secure than its predecessor, but it also opens new possibilities for hackers and other malicious parties. Automakers, they say, will have to prioritize 5G/6G security through encryption, verification, silicon security, and other techniques.

Perhaps the most obvious question as it relates to wireless and L5 vehicles, however, is how the cars will maintain a network connection and what happens when that signal is disrupted. Frank Schirrmeister, senior group director of solutions and ecosystems at Cadence, says a relevant concept is something akin to graceful degradation — maintaining functionality even when something happens to the network connection.

“It has to be, ‘I hope for the best but I plan for the worst,’” said Schirrmeister. “If I locally detect that the conditions changed in a way that was not predicted, then the car needs to have the ability to park itself. These things have to be thought out very carefully.”

One possible way forward is to utilize edge networks that are closer to the vehicles, which can minimize lost connections and latency.

“You can almost think of the cell towers as having compute stations on them,” said Paul Graykowski, senior technical marketing manager at Arteris IP. “It’s like a low-level cloud, where it’s doing some immediate processing there. And then once it gets information, it can pipe that right back down to the cars or then it can pipe it back up to the cloud. So you’ve got all these different levels that are going to have to come into play to make this work.”

Siemens’ Fritz says redundancy within the vehicle also will be an important part of ensuring safety. “The types of redundancy varies by company,” he said. “For some, there’s redundancy at the SoC level or even at the CPU, GPU, or NPU core level. That really is valuable, but it can’t support system-level redundancy. We need some cost-effective redundancy at the system level, and really the printed circuit board level. We’ve been working with companies on this whole concept of high-performance data coming across multi-gigabit automotive Ethernet, and that transitions into a redundant network on a printed circuit board. Now we have two subsystems operating simultaneously, checking each other’s inputs and output results.”

Adoption
There is little point in inventing a Level 5 car if no one wants to purchase or ride such a vehicle. Research shows that while many people are excited about a futuristic, hyperconnected lifestyle, they are also wary of ceding control of their cars to autonomous technology. A Cadence report released last year found that more than 70% of individuals surveyed preferred a car that allows a driver to have control over one that does not, and one that can have its autonomy disabled over one that cannot. The study also found drivers preferred to be advised on what to do, instead of the vehicle doing it automatically.


Fig. 1: A Cadence report showed consumer hesitance toward some autonomous vehicle capabilities among the majority of the 3,000 individuals surveyed. Source: Cadence

Overcoming consumer reluctance will require turning consumer’s worst fears into enthusiasm. Imagine being trapped in a self-driving car as it hurtles toward a dangerous situation with no ability to override the controls. Contrast that with the idea of watching a favorite movie during an easy commute to work and, with the click of a button, customizing the vehicle’s interior to look like the cockpit of a space shuttle or a cabana on a tropical island, depending on one’s mood.

Siemens’ Fritz says the possibilities enabled by 5G/6G that go beyond the basic function of the car are not to be underestimated. He says such features are part of what will convince consumers to actually use an L5 vehicle once the basic technology is in place and proven to be safe and functional. The first step is addressing the uneasiness that comes with relinquishing control of the vehicle. Fritz says that could be accomplished by a neural network that explains to the passenger why the car is making its decisions.

“Imagine yourself sitting in the backseat, no driver in the front, and you say, ‘Car, why are you not turning right?’” he said. “It responds, ‘I can’t turn right because there’s an ambulance coming. It’s okay, and it relieves the human anxiety, which is a real impediment to the adoption of these vehicles.”

He described the “easy allure” of pitching a car to consumers that, thanks to 5G and 6G, is not only safer than traditional vehicles, but can also stream multi-gigabits of videos or games or allow for a productive commute to work. The technology is alluring from the automaker’s perspective, as well, because customization and entertainment features are likely to be what automakers use to differentiate themselves once the basic L5 technology is in place. That will be increasing important as auto companies vie for share of a trillion-dollar-plus market.

“The real value down the road is in how the consumer can customize the vehicle to meet their needs,” said Fritz. “Auto OEMs are saying, ‘What we really want is to build something that is safe, secure, robust, and has the right compute power. But we also want our customers to be able to download apps from the app store, and safely and securely change the behavior of the vehicle.’ It could make it go faster in certain sports cars, or it could be as simple as making the internal vehicle act like the Starship Enterprise bridge. It could be separately streaming things for each passenger, like high-resolution video and real-time gaming while the vehicle is driving. Those are things you will differentiate on in the future. And those are just the most obvious. There are probably dozens of other incredible things that we haven’t even thought of yet.”



1 comments

Roger Stierman says:

The data speed of 5G connectivity may be a necessary condition, but is insufficient. Compare to the current failure rate for manual control (assume all accidents are unintended). Humans have an average accident (‘Failure’) rate of 168 FITs (i.e 168 accidents per 1,000,000,000 driving hours). But a Level 5 autonomous accident rate of 1 FIT will be publicized in nearly every news report. We need to rate the Level 5 systems by FITs and not by miles driven.

Leave a Reply


(Note: This name will be displayed publicly)