Last week, I began in part 1 by examining the difficulty of creating a new network system in cars when you can only network with people you randomly encounter on the road. I contend that nobody has had success in making a new networked technology when faced with this hurdle.

This has been compounded by the fact that the radio spectrum at 5.9ghz which was intended for use in short range communications (DSRC) from cars is going to be instead released as unlicenced spectrum, like the WiFi bands. I think this is a very good thing for the world, since unlicenced spectrum has generated an unprecedented radio revolution and been hugely beneficial for everybody.

But surprisingly it might be something good for car communications too. The people in the ITS community certainly don't think so. They're shocked, and see this as a massive setback. They've invested huge amounts of efforts and careers into the DSRC and V2V concepts, and see it all as being taken away or seriously impeded. But here's why it might be the best thing to ever happen to V2V.

The innovation in mobile devices and wireless protocols of the last 1-2 decades is a shining example to all technology. Compare today's mobile handsets with 10 years ago, when the Treo was just starting to make people think about smartphones. (Go back a couple more years and there weren't any smartphones at all.) Every year there are huge strides in hardware and software, and as a result, people are happily throwing away perfectly working phones every 2 years (or less) to get the latest, even without subsidies. Compare that to the electronics in cars. There is little in your car that wasn't planned many years ago, and usually nothing changes over the 15-20 year life of the car. Car vendors are just now toying with the idea of field upgrades and over-the-air upgrades.

Car vendors love to sell you fancy electronics for your central column. They can get thousands of dollars for the packages -- packages that often don't do as much as a $300 phone and get obsolete quickly. But customers have had enough, and are now forcing the vendors to give up on owning that online experience in the car and ceding it to the phone. They're even getting ready to cede their "telematics" (things like OnStar) to customer phones.

I propose this: Move all the connected car (V2V, V2I etc.) goals into the personal mobile device. Forget about the mandate in cars.

The car mandate would have started getting deployed late in this decade. And it would have been another decade before deployment got seriously useful, and another decade until deployment was over 90%. In that period, new developments would have made all the decisions of the 2010s wrong and obsolete. In that same period, personal mobile devices would have gone through a dozen complete generations of new technology. Can there be any debate about which approach would win? I've already heard the shouts -- the planned connected car applications would not work nearly as well in personal devices. They would not be on all the time. They would sometimes crash or die. They might be buried inside the car and not get a good radio signal path. With the band shared many packets will be lost to interference. They're more hackable. All of this is true. The solution would not be perfect. But the perfect is the enemy of the good here. That's because of the quadratic rule in networking.

This is not something most engineers are comfortable with. The idea of doing anything less than the best in a safety system is anathema. It is not always well understood how "inferior" systems can actually be better in the long run because of how they get adopted, and how they improve in the world of Moore's law.

A system deployed in 50% of phones that fails in 40% of them would be working for 30% of vehicles, and 9% of inter-vehicle pairings. A system deployed in 10% of cars that works in 99.9% of situations would only work in 1% of inter-vehicle pairings. And yes, deployment in phones really would follow an adoption curve that much better than deployment in cars -- even if the cars have a legal mandate and the phones are optional.

There's harsh competition in phones, and vendors are eager for an edge. "This phone might save your life on the road" is a pretty nice sales angle, even if at first the odds of it helping are low.

When I first floated this proposal, the FCC had not published its NPRM to take back the spectrum. My first proposal was that in fact the ITS community make a "deal" with the rest of the wireless community. The deal would have been to "give up" most of the spectrum that the ITS community "owned" in exchange for a rule that any mobile device vendor wanting to use that spectrum in their device -- as they all would -- had to implement some amount of car safety connectivity when the phone noticed it was plugged in to charging and travelling along the roads. My guess is they would have jumped at the chance.

The leverage for this deal may be gone, but it doesn't mean it could not happen. If it did, I predict that within just a few years every phone sold would be capable of doing V2V and other V2x and a few years later, almost every car driving would have such a phone in it. As that started happening, cars would actually catch up, and make use of the super inexpensive chips created by consumer pricing to put the protocols directly in the car, giving access to a superior antenna and other car internal data.

On the other hand, they might not have to. Over the same period, I expect to see a lot of progress in new cars in talking to their driver's phones. The phone will be able to receive all the dashboard data like speed and braking signals. It will be able to send priority alerts to the driver. (It's less likely it would be given permission to do emergency braking, but people don't even plan that for V2V due to security reasons.)

My latest phone has a coil for inductive charging. While a charger for my home would be nice, where I really want one is my car. In the car is where it's hard to hook up the phone to the charger. I want a little slot in my dash or armrest area where I can drop my phone, and it immediately gets inductive charging and a data link to not just the audio system. I think a lot of people want this, and once you have this, the phone is always powered in the car, and can thus afford the power budget to do always-on GPS and always on wireless data protocols.

The big thing missing is external antenna connectivity. There's nothing like a nice physical wire to an external antenna, but the convenience of just dropping a phone in a holding slot is pretty important. There are some techniques that can be used here, including allowing the car to do the analog RF part of things (this doesn't change too much with time) or making the phone slot be isolated or converting the RF signals to light or intermediate frequency signals within it. This is non-trivial as phones are interested in all sorts of bands -- GPS, cellular, FM, TV/ATSC, unlicenced bands, NFC, and probably more to come. Alternates include having the charging slot (which must be big enough for a tablet) be on the forward part of the dash or up in the roof above the mirror. You also need a fast data link from phone to car, to drive the car's touchscreen.

The use of personal devices would, as noted, drive down the cost of the hardware very quickly. There might be 60 million cars sold in a year around the world but more than a billion phones. And as this cost went down, cars would just start installing this hardware without a mandate because it's cheap, and it's useful due to all the other devices out there.

Another advantage is that pedestrians and cyclists could now easily carry V2V transponders with them, though such usage would consume battery power. However, they could listen at lower power and only transmit when it seems necessary. And this would not be a special device, but simply an app. Many cyclists, seniors, motorcyclists would run these, and parents would have their children run such apps if there was even a small chance of saving a child's life.

As I pointed out, doing V2V is hard. The value is difficult to demonstrate. But with the leg-up of mobile phones, I expect that people will dream up apps we have yet to think of. It seems counter-intuitive to some, but chaos and private innovation beats central planning most of the time, even in networked protocols and radio.

The final question is interference, something that's going to happen as the spectrum is opened up in any event. The trailblazing rule of the 2.4ghz band was that unlike the old bands, interference was the problem of the receiver as much as it was the problem of the transmitter. New generations of radio technology are getting better at this. If it becomes a real problem, it should be noted that modern devices mostly know where they are, and rules could in theory appear to restrict non-automotive uses of a small fraction of the band next to highways and signalized intersections. Enforcement would not be perfect but it might help.

None of this solves the other problems the V2V plans have. Different models of GPSs still give different results and make it hard for V2V systems to truly figure out if cars are in the same lane. Security is not a solved problem in V2V, and this makes it worse, but it also gives the opportunity for faster evolution of the security.

The potential for new protocols also helps robocars participate in a very positive way. Unlike other cars, robocars have a 360 degree picture of the world around them, and can broadcast that to other cars. Presuming you can trust it, that means that you can get a nearly complete and highly accurate picture of what's going on in the lanes without having transponders in the other vehicles. Get 10-15% penetration of robocars and now almost nobody has to transmit, including pedestrians -- and those who have receivers get real value from them. The robocars will uses these protocols for their own limited cooperation and it's not a big change to put these radios into robocars. Unlike regular cars, robocars will be very high-tech, with lots of computing power, and they will be getting upgraded all the time, more like computers than cars.

Phones, of course, also have their other radios, notably the cellular one. This provides data connectivity all the time, though it is through central servers and at higher latencies. Most ITS engineers disdain this channel because it's hard to do safety critical applications over such a channel, but in fact a lot of road situations work fine here. Traffic lights don't generally change by surprise, they have a few seconds warning even when doing a change for an ambulance, and so their data can often flow over this network. And stalled cars and accidents do not have to be reported within milliseconds, so the wide network is good for that too. This network already exists, which is an unimaginably great advantage over any proposed network that doesn't yet exist.

In part three -- using broadcast data, I will detail another potential use for a network that already exists.