At CES this year, companies used the Las Vegas trade show to showcase flashy demonstrations of autonomous vehicles. Mercedes-Benz shut down portions of the Strip to test out an autonomous concept car. Ford’s CEO told an audience that his company would test a platform for providing autonomous deliveries or rides in the future, working with companies like Domino’s, Lyft and Postmates (Lyft, for its part, also offered self-driving rides to attendees at CES last week).

Marcus Welz, CEO of Siemens Intelligent Traffic Systems, was at the conference for a related but different reason. He was there to demonstrate the infrastructure needed to make driverless cars safe and reliable. As automakers work to create a mass-market driverless vehicle, cities and states are being forced to grapple with what that means for existing road infrastructure.

During CES, Siemens announced a demonstration with the City of Las Vegas to install an infrastructure system downtown that sends and receives signals from cars. Even before the switch to driverless vehicles, Welz said cities need to be thinking about this type of infrastructure. Human-operated cars are becoming increasingly connected, too. The infrastructure, which could ultimately enable autonomous vehicles, allows all these connected vehicles to communicate with each other and the city’s road network through a secure wireless signal.

“We believe having self-driving vehicles on the road without connected vehicle technology on the infrastructure side is like flying an airplane without a radar,” Welz said. “It may work but it won't be safe.”

The Nevada Independent talked to Welz about driverless cars, the future of road infrastructure and the Las Vegas project on Casino Center Boulevard between Bonneville and Clark Avenues.

How did the Las Vegas demonstration come about?

First of all, road traffic infrastructure is currently changing dramatically. We will [still] have all of the traditional equipment such as signs and traffic lights. But they will all become more and more part of an integrated network. And the cars themselves will become traffic data generators and information prosumers. The technology that we are talking about and demonstrating here in Las Vegas is called connected vehicle technology. It is not only the cars themselves that are getting smarter but also the infrastructure on the road. This technology enables infrastructure such as road corridors, traffic signals, street lights and curves to communicate directly with the cars. The car is transmitting its location, speed, direction, latitude and longitude 10 times a second to the infrastructure. And the infrastructure is able to respond to that in real time, giving warnings and recommendations to the driver to increase safety or improve traffic flow.

Ten times a second. That is mind-boggling. What do you need to transmit those signals?

The communication channel between the car and the infrastructure is DSRC (Dedicated Short-Range Communication). There is a 5.9 GHz band for this communication technology, which is only for safety applications in transportation. This technology is currently the only technology that is able to follow all of the safety requirements, cybersecurity requirements and also the latency [requirements]. Today cellular communication doesn’t have the required latency and accuracy in order to do that. Once 5G is ready to launch, that’s a technology which will probably be considered in parallel to DSRC.

Is DSRC widely deployed or is that something you had to set up with the city?

There has been an alignment between all of the car manufacturers and the infrastructure providers to use DSRC as the standard communication channel… The DSRC is nothing specific for Las Vegas. The first city that has put this technology in place on a large scale was Tampa.

But Las Vegas deployed it?

At Casino Center Boulevard.

And it makes your project possible?

Exactly. Our focus is mainly on the infrastructure side… We believe having self-driving vehicles on the road without connected vehicle technology on the infrastructure side is like flying an airplane without a radar. It may work but it won't be safe. Connected vehicle technology is a fully redundant system that gives a self-driving vehicle the information [it needs], in order, for instance, to stop at a red light or be at the right level of speed to traverse safely through the city. We consider the infrastructure as an enabler for the innovations that are happening on the vehicle itself.

When you approach policymakers in cities like Las Vegas, what are their biggest reservations bringing in an AV project or even just a demonstration?

Currently, cities invest a lot of money in traffic but more into traditional technologies. They invest a lot of money to put sensors on the road — dig up the road and put a sensor in place in order to detect the cars. In the future, this won't be necessary. Whether it is a self-driving vehicle or a connected vehicle, the cars will tell us where we are.… The city can actually shift the investments from what they invest in today into those new technologies.

I was here when Lyft and Uber came. They faced resistance from the taxi industry. A lot of that had to do with employment. Given that there won't be any drivers for AV, is that going to be an issue for trucking and ground transportation, which is huge in Las Vegas?

Ultimately, I believe that any transformation will be done over time. It won't be that within a three-year period, all the traditional mobility services, like public transit, will go away including the driver. But it is a challenge for future mobility services to make sure job creation and equity in transportation can still be maintained. Solely focused on the connected car and the connected infrastructure side, there is nothing that really replaces today's jobs or infrastructure. It is more the self-driving vehicle by itself that might have an impact on busses and public transit agencies.

Your demo is here for six months?

No, it is here indefinitely. It is part of the infrastructure now.

How long can people expect to wait before this is a ubiquitous experience?

Consider that a traditional product life-cycle of a car is maybe 10 years. Today the population of cars on the road that already have those capabilities to talk to the infrastructure is less than five percent… But whenever a car, whether this is a privately owned car or a fleet, is being renewed, then the population of vehicles that (is) able to communicate with infrastructure will increase. We believe that in 10 years from now, it should be at a maybe 80 percent saturation level.

Your demo exists with regular traffic?

Exactly. Yes, it is here working in regular traffic.

What are the difficulties of pulling that off with cars that are unpredictable?

It is pretty simple to deploy. In order to make one location — an intersection, a corridor, a curve, a road-section — able to communicate, it is done in less than four hours…. The biggest challenge is that we have [to have] a very accurate topology of the intersection. So the map file…needs to be 100 percent accurate. If it communicates to the car, it needs to know exactly whether the car is on the middle lane, the right lane or the left lane. The infrastructure side is also adjusting [for] the location of the vehicle…. To have those map files accurate and help [get] the geo-positioning of the car via the infrastructure, this requires a bit of engineering work.

How big is that file?

I have no clue actually.

Let's say the AV vehicle is going and all of the sudden some guy swerves into the lane or goes through a red light. How does the vehicle know how to react? How does that work within the infrastructure system you have?

What we are doing from the infrastructure side is providing all that information in real time to the vehicle…. An automated vehicle will then make a decision, for instance, to brake or to re-route. This type of decision-making is fully done by the robot in the vehicle itself.

How much information are you transmitting in a second, a minute, a day?

I don't know exactly.

It must be a lot.

It is a lot. We package this data — the vehicle precision, the speed and the direction. And this is being communicated 10 times a second. With this latency of 10 times a second, we are able to give real-time warnings…. At the end of the [day], these are safety applications. You cannot have a one-second latency if you are approaching an intersection and want to know whether you are about to go through a red light. You need to know in less than a second. Otherwise, this information is useless. So the latency is important.

Is there a machine learning component to this?

There is actually an [artificial intelligence] component, especially in the area of changing the patterns of the traffic signals. If you know the real-time traffic situation from all of the vehicles, then you can really optimize a traffic system with that data available. Instead of running on a fixed mode, [the traffic signal] is running on an adaptive mode…. The speed of the communication will ensure that this algorithm is optimized by itself every second.

Is there anything you want to add?

Besides all of these advantages on safety, it also gives a city or even a complete state an overview of the real-time traffic situation… And then based on that, [they can make] a decision on changing the traffic patterns on a traffic signal [or] changing a speed limit on a highway.

This interview has been edited for length and clarity.