HAVstory: What’s at the core of the rivalry between DSRC and C-V2X?
Kenney: V2X technology is cooperative, and that means we need to have a standard so that the bits can get through – so that my message can get to you, and you understand it. That means from a data networking point of view, protocols have to line up, all the way up and down the stack.
That’s not fundamentally unique to V2X, but the thing that is unique to V2X is that it’s a direct, ad hoc type of communication that, as I said in the panel, means I’m not going through a base station.
With a cellular network, or a WiFi network – Let’s take the WiFi example because I don’t want to attack cellular, if they write an 802.11a, then n, then 11ac, and now they’re going to write an ax standard – If they publish a new standard and the access point can support it, I don’t have to change my equipment, because that access point is still going to be able to understand me, and, just as important, the other devices on the wireless LAN are going to be able to share the channel with me. I may not be able to decode their bits, but at least I can co-exist with them on a channel-sharing basis.
That means they’re going to send a signal at the beginning of their transmission that I can understand as the start of a transmission and that tells me to back off. Likewise, they can understand mine. So that community, and the cellular community as well, are used to coming out with new versions every 15, 18 months and the access points or the base stations handle all of that messiness.
We don’t have the benefit of that in V2X. A lot of the V2X stuff is ad hoc networking that comes out of the military. Imagine a battlefield where you’ve gone to the trouble of equipping all your soldiers, and all your tanks and equipment, and some platoon shows up and says, “I’ve got new equipment.” That defeats the whole purpose of setting up the network.
That’s the baseline. The other thing that comes out of that is that our deployment model is different. Usually, a company like Toyota evaluates a new radar, or airbag, and we control that entirely. We can make the cost-benefit analysis and decide if that’s right for our customers. But with this technology we have to make guesses about what everyone else is going to do.
We’ve done that, and we think [DSRC] is going to be there. GM’s already there, and the state DOTs are coming online, and we think other automakers will come too, and we want to encourage that.
But it’s fundamentally harder, and we see a lot of the smaller players, their instinct is to hang back and wait. First, they saw this WiFi spectrum issue arise five years ago, that cast significant doubt in some people’s minds about whether we were really going to be able to execute the mission of DSRC and now there’s a potential new risk: Maybe it’s not going to be DSRC, maybe it’s going to be some other technology.
That just impedes deployment, and to be honest with you, from the Cellular V2X side, I think part of the strategy has to be to impede DSRC deployment.
I want to be careful, because I don’t want to come across as critical. It’s one thing to say... and some of our colleagues on the C-V2X side are saying, “We understand DSRC and believe in it, we just want to evaluate this other thing.” It’s different to say that than it is to naysay DSRC or say that it can’t meet the mission.
HAVstory: What are the compelling arguments for C-V2X?
Kenney: Not much. There are some aspects of their specific protocol that take advantage of their, let’s say, newer concepts that aren’t part of DSRC. As an example the type of error control coding that they apply, and the actual waveform structure have some inherent advantages.
Let me back up; when I think about the big question, I see lots of different dimensions. Performance comparison is just one of them. It’s a valid one, and they talk about that a lot. In a performance comparison you might expect to see some advantages coming from things like the coding and waveform; you might also expect to see some disadvantages that we’re starting to uncover.
Let me explain just one: in DSRC – I’m a radio – I have a packet to send, the first thing I do is, I listen to see if someone else is already sending. There’s a very specific protocol for doing this listening and if the channel’s idle I can go ahead. If it’s not idle then I wait and when that transmission ends, I execute a set of steps to try to avoid jumping in when someone else is about to jump in.
It’s like being on a conference call, and you know other people have something to say and they’re trying to decide how long to wait. Occasionally when I execute that, it’s going to happen at the same time as someone else.
We call that a collision; not a car collision but a packet collision. Those are random events and they happen independent of the transmission. On the C-V2X side, they have certain pre-defined time slots that they think of as opportunities or candidates and they listen for a second and see which of those are in use. And they make an assumption that everyone is using these things periodically, which is not a good assumption to make in DSRC; some of our applications transmit periodically, but some do not.
They assume everything is periodic, if they can identify a time slot that hasn’t been used, they can select it. If you and I do that at the same time, there will be another packet collision. The difference is, they don’t make that decision independently for each transmission. Once they make that decision, they use it again and again. So if you and I are colliding, we’re going to collide again and again. I think on average 50 times. That’s a significant impact on performance. If our Basic Safety Messages aren’t getting through for 50 transmissions, other cars aren’t even going to know we’re there, let alone avoid collisions.
So the difference between an isolated, independent packet collision with DSRC, versus a persistent collision that can happen on the LTE V2X hasn’t really been studied yet. I haven’t seen any evidence that the C-V2X people are interested in studying that; I hope they will. But it’s been brought up. There are three or four other problems like that, that have been brought up by companies like Autotalks and NXP that are in the silicon business, as concerns.
Where does that take us in a technology comparison? Well there are going to be parts where LTE V2X is better, there are going to be parts where it’s worse. What is the net of all that? I don’t think we know, but I suspect it’s marginal.
Out of all the different dimensions, the technical or performance dimension is probably not the determinative one. We have to think about cost, and evolution, and readiness for deployment. And I think they make arguments that they have cost advantages that I don’t think are probably true. They make arguments about being able to evolve that ignore the reality of evolution in an ad hoc network.
On the evolution front, it’s probably a wash; they probably both have challenges, and IEEE and 3GPP [Third Generation Partnership Project] are both working on them. On the cost side, they’re playing on the idea that many, many cars are going to be equipped with LTE modems to communicate with the cloud for OnStar kinds of services. And that’s true; we’re doing that with all our cars. They want to claim that if you have an LTE modem in your car, then having LTE V2X is essentially free.
That’s fundamentally misunderstanding what the costs are. It’s different hardware, even if you put it on the same chip; it’s a different antenna; cabling, there’s a lot costs there; it’s running on different spectrum – It’s really two radios that share very little synergy – and if you believed in those synergies you could make the same argument for synergizing DSRC with a WiFi connection in every car, because cars are going to have WiFi too.
I would argue that 802.11p is more synergistic with the WiFi family than LTE V2X is with the LTE family, and here’s why: In the LTE world we have user equipment and base stations and most of the intelligence lives in the base station. It takes care of a lot of that management stuff and the user equipment can be simple and cheap.
In the LTE V2X world, now you have two pieces of user equipment. That means the intelligence from the base station has to be pushed out. That’s adding complexity. For example, they have to now do that thing I was describing, of listening, finding candidates; they never had to do that in LTE because the base station does that for them, but now they have to take that on, they’re adding functionality, complexity, and cost.
On the other hand, when we wrote the 802.11p standard, we took the 802.11 standard, and took things out. We took functions out. In 802.11 a lot of the complexity has to do with associating with an access point and authenticating yourself. There’s a whole handshaking process that happens when we turn our devices on and connect to an access point. We took that out; there are no access points in DSRC. Everything is ad hoc, device-to-device, so it’s easy to put 11p into your 802.11 chipset.
I’m not a great believer that there’s synergy in either direction but I think that if anything the cost advantages are going to be on the DSRC side.
HAVstory: On the face of it, companies like Qualcomm – a chipmaker – can make chips for either DSRC or C-V2X. Why do you think they are so committed to C-V2X.
Kenney: You’d have to ask them. There’s speculation that they will profit more from one than the other. They have a DSRC product, they have an LTE V2X product. What those costs will be, I don’t know. There are people who talk about [Qualcomm’s] intellectual property, but I haven’t studied that.
It is true that in the 3GPP world, it’s more common to standardize things that are covered by intellectual property rights of the company. In the IEEE it’s a little less common to do that. In the DSRC world, we made an effort to avoid patenting our ideas and we were pretty successful with that. The rules are that if something’s being proposed, to be essential, is covered by intellectual property, you have to disclose that, and there were very few instances of it.
HAVstory: Some people think that in the end, cars will use a tandem system, relying on DSRC for BSMs and also having a 5G connection. Is that plausible?
Kenney: I think the more likely thing is that, if we run into things in the future that DSRC can’t handle well enough, we will turn – not to something completely different and incompatible, like cellular V2X – we will turn to an enhanced version of DSRC. In IEEE 802.11, they’ve started a study group called ‘Next-generation V2X’. They had their first meeting in May, which I attended, they’re having their next meeting in July. They’re working right now to define their scope and mission. They’re in the study group phase, they’ll move to the task group phase by the end of the year.
What they’re probably going to do is create an amendment that includes the same kind of new, extra features that we talked about for C-V2X; better error-control coding, maybe some improvement to the wave form structure, some other improvements for multi-input/multi-output antennas. They’re going to create an amendment that will become an option to employ, if needed.
My guidance to them was, Please understand that the “if needed” part is really critical, because we don’t want to do that just for the sake of better technology, because to do so might disenfranchise the existing cars.
Let’s say a new application comes along a few years after a car’s on the market. It’s entirely possible that we can add software to that car, which will enable it to run that application, if it runs on 802.11p, because we’re probably not going to change the hardware; we’re not going to change those lower layers of the protocol stack. Let’s try to support applications on the most broadly deployed technology that can support them. That probably will be 802.11p for a number of years.
The advantage of an advanced version of IEEE, compared to C-V2X is, they can operate in the same channel, because of what I was describing earlier; they’re going to employ these techniques so that they can tell when somebody’s already on the channel, so they can start their signals the same way, so that the 802.11p device knows it has to back off; they can co-exist. Furthermore, that advanced IEEE device is still going to be able to communicate with those legacy cars over the core set of applications like the exchange of Basic Safety Messages, that we’re already in deployment on. Those newer cars will be able to interact with the whole population of 802.11 cars.
That evolution makes more sense to me than evolving from an 802.11 system to a C-V2X system because they can’t even be in the same channel. So then, you have to split the spectrum, which is just a terrible idea; it’s spectrally inefficient. It implies that some people are going to send BSMs on channel x using technology a, and others are going to send them on channel y using technology b. That’s a terrible idea and I hope the FCC sees that.
HAVstory: Is there any reason that C-V2X has to operate in the 5.9 spectrum?
Kenney: It doesn’t need to. The thing is, the 5.9 spectrum is free public spectrum. There’s no business case for operating this in an auctioned band that we pay billions of dollars for. No operator wants that. There would have to be some new spectrum that could be operated for free, and in fact the 5GAA has made a request of exactly that nature to the European spectrum regulators and I think they’ve made a similar request of the FCC. Both the FCC and CEPT (the European spectrum regulators) are in parallel pursuing opening up the 6 gigahertz spectrum for something, and the WiFi people say it’s for WiFi and the ITS community could say we need some of that, too. The 5GAA could say, Give us another 100 and if that happened you could make the case that, Let’s use that for some other technology. I’m still not sure that makes the most sense.
One other idea that’s important to understand is, we have seven channels but we’re not going to put seven radios in our car. We’re going to put two radios in our car. One is going to be constantly on the lowest channel where the BSMs are and it’s never going to leave. The other one is going to hop around the other six channels – it’s going to channel switch – in response to service announcement. So if we fragment the spectrum and say, These three are for one technology and these three are for another, you defeat that whole channel-switching idea.
Even if we could augment our 75 MHz with another 100 MHz, my preference would still be to do channel switching across those bands. I’d still prefer that it was one technology, but at least it would take some of the pressure off the idea that introducing a new technology would disrupt the current [DSRC] plan, because the current plan is to use 75. The current plan, to come in and share spectrum, could delay or even kill [DSRC].
If instead they got a new band, it would not disrupt DSRC.
HAVstory: For a decade or more, every car company agreed that DSRC was the way forward for V2V. What explains the change of heart on the part of Ford, Daimler, & BMW?
Kenney: A few of those companies have become enamored with C-V2X. I don’t know how to explain that, except to say that there are clever people in these stakeholder companies who are doing a great job of selling these ideas, and they’re selling them on the basis of lower cost, or better evolution strategy; which I’ve explained that I don’t believe, or higher performance, which I think will be negligible, but they’re being successful with a few of these companies – more the luxury car companies, and Ford.
HAVstory: We’ve seen user data become a valuable commodity. Does that explain some of the desirability of C-V2X?
Kenney: You can capture data on an aggregated basis, but there’s not individually identifiable data in the DSRC world.
Are we seeing data being monetized? I don’t know that it would be easier to monetize user data in the C-V2X world; they plan to use similar privacy ideas, the same basic safety messages and other core messages. I’m not aware of a way that data would be capture
One thing to be aware of is that kind of protocol competition is not only happening here in the U.S. It’s also happening in Europe. My impression, one big difference is our FCC requirement is not technology-neutral. In Europe they have ITS spectrum but there’s no specific technology. The industry has been moving towards DSRC – what they call ITS G5 – it’s all the same companies want to do the same thing. They’re having the same fight there, but without the built in technology regulation that we have.
My impression is that the European Commission is moving towards a tacit recognition that DSRC is the incumbent technology that people should work towards. They do emphasize interoperability but at the same time technological neutrality, which are competing ideals. China’s also having this discussion, but they obviously approach regulation in a different way, and it seems they will go with cellular V2X. So, we’ll sell cars in China and we’ll use C-V2X and we’ll make it work.