njkayaker

"5G" is mostly a buzzword in that article (like "blockchain" is). All the stuff that article talks about can be done easily without 5G (or blockchain).

"5G" is just a faster version of already-existing cell data-transfer networks. It's (generally) shorter range (so, it could be used to determine location better than other cell networks).

GPS just lets a receiver determine its location (and nothing else). Since the purpose of GPS is is nothing like that of 5G, it doesn't make sense to say it "isn't as powerful" as a bidirectional data-transfer network (5G).

GPS (with augmentation) is still better for location than anything else.

Milton Keynes

I remember when I was a kid in the 70's reading a book from the 50's saying "You will go to the moon/" I think the authors would be shocked to know that, while we have been to the moon, in the 21st century we still don't have any moon bases yet.

livedarklions

I'm going to Walgreens to have hundreds of tests performed using only a single drop of blood.

PeteHski

The biggest shock is that we haven't been back to the moon for exactly 50 years and counting.

Gene Cernan's last words on the moon 1972: "as I take man's last step from the surface, back home for some time to come—but we believe not too long into the future—I'd like to just (say) what I believe history will record: that America's challenge of today has forged man's destiny of tomorrow

DangerousDanR

I was involved in a collision avoidance radar for automobiles project in the early 1980s.

There were two conclusions at that time:
1) you would need a truck and a lot of power to make it work in 1983. That problem has largely been solved.
2) determining the intent of a vehicle parked by the side of the road is impossible without vehicle to vehicle telemetry. This is still true, and the amount of information you need to share to solve this issue is not trivial.

PeteHski

While this is the obvious advantage of V2V communication, it is not essential for safety providing the parked vehicle is able to effectively monitor its own surroundings before making a move. At the moment this is largely down to the judgement of a human driver (with the inevitable errors) but smarter cars are reducing the manual driver workload. At least in terms of alerting the driver to what is happening around their car.

I think the challenges of vehicle-to-vehicle telemetry are even bigger than the challenges of making individual vehicles self-sufficient. The latter can at least co-exist on roads with "dumb" cars, which is the situation we are seeing on roads today, with a small percentage of vehicles able to detect and react in a limited way to their surroundings, but most are relying 100% on human senses and judgement. Even if all vehicles were fitted with a basic electronic V2V beacon, I don't think it would add very much to what semi-autonmous cars are already able to "see" with their various sensors.

DangerousDanR

I don't think we disagree all that much. The V2V would need to be fairly robust and bidirectional to add significant value.

The big challenge was this: if our system detected a vehicle by the side of the road with the engine running (we also did FLIR so we could see the heat signature), was it going to pull out in front of us such that we would not be able to prevent a collision.

Our answer was that a manually controlled vehicle could easily do that, even if we saw the exhaust heat signature before the parked vehicle began to move.

So do we slow to a speed where we could avoid the collision? That is going to sell really well.

Now let's assume that there is no heat signature? We also had access to an electric vehicle test bed, so we knew that was a possibility.

Also, how do we handle vehicles blocked from the view of our system? Think of the situation where there is a truck parked with the flashers going or in an alley blocked from view by a building.

If we know that there is a potential vehicle that can enter the road we can maybe avoid it if it moves in an unsafe manner. But probably not.

And what about adverse conditions, like snow or RF emissions from somewhere. Both vision and RADAR systems can go "blind."

The parked vehicle needed to have a system to prevent it from entering traffic and in 1983-84 we couldn't rely on that.

And is there a manual speed override on our vehicle? If yes, then what happens if the operator requests a speed increase just before we reach the safe vone for the parked vehicle?

Just "I am here moving in this vector" (including speed) might not be enough to solve the problem.

Universal V2V is a very big challenge. We couldn't do it then. Data rates are much higher now, but I believe it is still a big challenge.

Lots of stuff in a vehicle happens today at CAN bus speed, so even once a vehicle receives a telemetry message there is still going to be a very small delay before it can act on the information.

As long as there is the possibility of human interaction and override of the system, or human control of other vehicles, we could not meet the requirements of our legal team.

I really want someone to get a true self driving car out before I am no longer a safe driver, but I don't see it happening. But 20 years is a long time so I can hope I am wrong.

PeteHski

I'm doubtful too. The step from semi-autonomy to true self-driving is much bigger than the likes of Elon Musk first thought. Having said that I'm pretty happy with Tesla Autopilot for cruising along motorways, which is a relatively simple case. In the summer I drove from the UK down to Southern France and about 90% of that journey was on AP as it was pretty much all motorway until reaching the Alps. If it wasn't for the constant steering nags it would have required very little driver input. To be honest I'm not all that fussed about true self-drive in all scenarios, but the challenge is certainly interesting.

One thing that bugs me a little about the current generation of semi-autonomous systems is how easy they are to over-ride. I understand the various reasons for this, but I think we are getting to the stage where it should be impossible to deliberately/accidentally run over a cyclist in our path. Almost without exception, my Tesla is fully aware of an upcoming cyclist and will brake automatically for them in AP. But I'm free to either switch off the system or simply over-ride the throttle and steering. Maybe it's time for these systems to over-ride the driver in obvious scenarios like this when they do something stupid. I know some systems will prevent the vehicle from hitting objects while parking, but that seems to be about the limit when it comes to the vehicle over-riding driver input.