AT&T is required…
First of all why you think what ATT is doing is relevant? Is there anyone developing any technology with ATT for AV’s?
How do you thing AV’s communicate.
Who do you think sits on all of committees that build the standards?
AT&T is old and slow. AT&T understands the value of the data. How much more value do you think the cutting edge guys place on the data?
The cellular network is not required for instant communications. It is the backbone that lets the AV talk to each other. It is the way that an AV 22 second in line for a slow down on the interstate can know when the first AV puts on its brakes.
Think about it. You get a sudden slow down on the interstate today, the first guy hits his brakes, the second guy uses his cushion, the third guy and so one. By the time it gets to the 15 guy, somebody is going to get a visit to the body shop. But, if even 1 in 4 cars know the instant the first car hits his brakes, the whole line of traffic has the chance to avoid a crash.
Also, to you thoughts on availability, with 5 G, there is some work going on to use the HSS or HLC (They changed the name between 3G and LTE and don’t remember which is which.) The cellular system can authenticate a device, and give it a token, then allow devices to build authenticated adhoc networks. In this way, rather than send a signal to a tower, through fiber, a few routers and back, only the token gets sent, then the two devices can speak directly and know that they are speaking to a secured, or authentic device. When this happens the round trip time can be drastically reduced, even though the data through put will be slow. Think 0.1 megabits per second with a 1 ms ping time.
If you are only passing data that looks like this:
Temp
Humidity
Barameter
Traction control flag
Anti skid flag
Current threats = 8
G meter vector
Threat 0 = [lat, lon, timestamp, speed, course, alt, horizacc] = poslog(m)
Threat 1 = [lat, lon, timestamp, speed, course, alt, horizacc] = poslog(m)
Threat 2 = [lat, lon, timestamp, speed, course, alt, horizacc] = poslog(m)
Threat 3 = [lat, lon, timestamp, speed, course, alt, horizacc] = poslog(m)
Threat 4 = [lat, lon, timestamp, speed, course, alt, horizacc] = poslog(m)
Threat 5 = [lat, lon, timestamp, speed, course, alt, horizacc] = poslog(m)
Threat 6 = [lat, lon, timestamp, speed, course, alt, horizacc] = poslog(m)
Threat 7 = [lat, lon, timestamp, speed, course, alt, horizacc] = poslog(m)
Threat 8 = [lat, lon, timestamp, speed, course, alt, horizacc] = poslog(m)
With proper coding, this data would barely take up a single Ethernet frame.
What is more, once a AV identifies it self, the Threat and number gets replaced with a authorized token and each AV redistributes the info about its self and every other AV, much like the ARP protocol does for Address resolution in networking. The difference is, if you use the current cellular network authentication, you get a much more trusted system.
Very quickly the AV’s can assess threats and classify them into AV and non AV threats. Using algorithms like Bayes rule, with in just a few seconds it can make a fairly reliable assume that as to what the threat is like. Is it a car, a train, a plane, Underdog? A threat moving outside the roadway at sub 2 miles per hour with a volume less than 2 cubic meters can get a high probability that it is not a car.
Within a few months the combined AI will be able to determine that the thing moving at 2 miles per hour outside the road way is probably a human and then add it into its system.
And yes AT&T is absolutely working in open organizations to make the network work like this.
Cheers
Qazulight