Forums

Tesla & 5G

Tesla & 5G

Do you think 5G is required in order for Tesla to achieve full self driving feature? When do you think Tesla cars will equip with 5G? Any speculation on new upcoming vehicles like Model Y or the Pickup truck will have 5G capabilities standard?

jordanrichard | January 8, 2019

Cars will not and should not use cell signals to enable their cars to self drive.

reed_lewis | January 8, 2019

No. The car itself needs to be able to 'self drive' without any internet data.

Plus at 5-10 mbps, the connection speed is fast enough.

Darthamerica | January 8, 2019

Required, no. Helpful, yes. Especially when real time traffic and map updates are part of the system. With a higher speed connection there are other features that could be added as well. For example the car could "see" potholes or other road hazards such as road construction. If the car can push this data to Tesla cars Navigate on Autopilot could do things like avoid a particular lane or manuever to a faster lane to avoid slowdowns. The cars could also communicate with each other. But these are things that will take 5 to 10 years before I expect to see it.

reed_lewis | January 8, 2019

The current LTE speeds of 5-10 mbps are more than adequate for these purposes. The only reason that a higher speed connection would be needed is for live broadcast video.

Never underestimate the power of tokenizing the data. Pothole in an area? Just send the Lat and Long and location of the pothole.

And the throughput of the data is less important than the latency of the data. 5G will do very little to improve latency from what I have read.

Darthamerica | January 8, 2019

Reed it makes a difference. Especially in a moving vehicle where your interference is unpredictable. The faster you push bits OTA the more robust you are. I'm not saying LTE isn't good enough but 5G would be a tangible gain in performance.

jordanrichard | January 8, 2019

Unless the self driving car is connected to/receive a satellite signal for all of the pertinent data, it won't work.

This is like comparing insurance rates, there are far too many factors/moving parts. The source of data that the car(s) needs has to come form something that is not dependent on cell signals that can be cut off by a power outage or a dead spot in coverage.

Perhaps that is why Space X just sent up a crap load of small satellites.......

Darthamerica | January 8, 2019

Jordan all base self driving functions will need to be on board. The improved radios are just for additional features that require connectivity.

jerrykham | January 8, 2019

Realize that there is a very real chance that legislators won't allow FSD without V2I or V2V communications. I agree that LTE is sufficient for the car to get traffic updates and change routing based on accidents, etc. We will have to see what requirements various regulatory bodies come up with as regards to requirements for FSD. Clearly the actual decisions need to be made in vehicle and not depend on a connection to the internet. There will likely be provisions about redundancy of components. Legacy automakers have already started harping on that and showing that they are building it and that Tesla didn't - those guys have lots of money and lobbyists so very like those provisions will make it in - from what I understand AP 3 hardware builds computer redundancy in so we will see.

5G is unlikely to have any bearing on it. I would imagine since 5G has such short range (mmWave doesn't make it through obstructions well at all) that it may be pretty hard to get working reliably in a fast moving vehicle in the short term. A lot more tower switching that you do today with LTE. And probably a lot of "now you are on 5G, now LTE, now 5G".

jordanrichard | January 8, 2019

Darth, in order for a self driving car to know traffic conditions like surrounding traffic and V2V communication, it needs that information from an external source.

Darthamerica | January 8, 2019

Yes I understand. That's where the better radio can help. Another use for it with FSD will be to allow for entertainment while the vehicle drives. We've got huge 17" and 15" monitors so there's a lot of potential as the technology improves.

reed_lewis | January 9, 2019

The interesting thing about 5G right now is that there is no spec for true 5G out there. AT&T is coming out with a 'puck' to do their '5G', and Verizon only has fixed 5G service for home. The main issue is that both of these companies are still limiting the amount of data that can be used before throttling, and with the 5G service you would be able to use up the data allotment within 30 minutes or less.

And there is a major difference between latency and throughput. On a cellular connection, faster does not always mean less latency. Just because you can push a lot of bits in a second does not mean that the first few bits will transfer faster.

And I concur with @jerrykham. The car needs to drive itself using its own built in system. There is no way that any product would be approved if it relied on a reliable internet connection.

Darthamerica | January 9, 2019

More insight and why it's likely 5G will play a crucial role in Tesla's FSD future.

https://youtu.be/dA3p0FK4JxM

https://youtu.be/GEx_d0SjvS0

jordanrichard | January 9, 2019

Got it Darth, it may be crucial, but it can't be dependent on it to function. It's like cruise control. Cruise control is crucial to help alleviate fatigue from long drives, but the car can function without it should it fail.

Look at the recent firs in CA where the town of Paradise was wiped out, along with the cell towers. How is a FSD car going to function within that cell area if it is required for the car to function.......

Darthamerica | January 9, 2019

Jordan I agree with you and mentioned that the car will need to have all base functions on board. This is necessary in case of coms loss. But it would likely limit the car's capability to be closer to what enhanced autopilot is today. Self Driving on open roads is almost a general purpose AI problem. I don't think that technology is ready yet in any car.

It's likely that regulations will do to self driving what speed limits do for roads. On certain pre-approved public roads where a car is connected, can be monitored and meets certain standards FSD will be enabled. On unapproved roads certain functions may be disabled for safety and liability reasons.

jordanrichard | January 9, 2019

Yes, but road and traffic conditions up ahead, for the purposes of navigating can’t be “on board”. How is a FSD car to know that the road up and ahead is closed due to an accident? Right now the traffic based navigation in any car is reliant on third party information transmitted via the cell signal.

TeslaTap.com | January 9, 2019

@Darth "I don't think that technology is ready yet in any car."

I see it every day. Google runs a fleet of self driving cars (although with a backup driver so far) in our area. I also see a number of other unidentified FSD cars daily - likely Apple, Cruise Automation and a few other companies located here. Now none of these are available for consumers to purchase, and it appears the costs to add just the hardware is crazy high right now - but they are driving now. I'd say Waymo (Google) appears the best operationally. It's unclear if or when it might become economical to sell.

All of the technology I see being designed for V2V and V2I communication uses a specific slice of bandwidth and does not depend on any cell like operation. Less items in the path greatly increases the service reliability and greatly reduces latency.

Let's take V2V. Car 1 communicates directly to all cars nearby. It only needs to work on short distances too, as only cars nearby are of any concern. The systems are also designed to provide total privacy (it's a really complex problem, but already solved).

Now try and do this via Cell. The car has to have a cell signal, it has to communicate with some server somewhere via the internet. That server then has to resend the information to all vehicles nearby. This means total tracking of every vehicle so it knows which cars to send the communications too, a huge privacy issue. Then all the cars nearby must receive that cell signal, requiring a huge amount of connections. I really don't see how it could be workable on so many levels. Bandwidth is the least of the issues with this method.

Darthamerica | January 9, 2019

@TeslaTap there's nothing ready for market yet, just prototypes and test miles. Bandwidth will be a huge issue with many hundreds of thousands of vehicles then millions. Spectrum is as valuable as gold these days...

reed_lewis | January 10, 2019

I concur with you @TeslaTap The only bandwidth that matters is in the local area when cars are communicating with each other using direct car to car communication. Using a central cell connection to do that would introduce delay that would be unacceptable.

And information on road conditions, etc. would take so little bandwidth that 3G could adequately fill that need.

Think of it this way. There are two types of data.

The first is data that a central location would need to supply because it is information that is needed but is too far away for local transmission. This is typically low bandwidth info which does not need millisecond latency. This might include road conditions, traffic, closures, damage to the road, etc. That data is not going to change multiple times per second.

The second is data that is not needed by a central storage system, but needs to be millisecond accuracy. This would include other car locations, stoplight states, etc. This would be handled by a local peer to peer multicast type network. A sort of wireless ethernet system where any data that is broadcast is received by all other cars in the immediate area (say a quarter mile or less depending on traffic conditions). In this case the issue of bandwidth is moot because the domain of coverage is very small.

But 5G does not meet either of these needs. So I stand by my original premise that 5G is not needed for any in car automoney.

Earl and Nagin ... | January 10, 2019

This V2V and V2I issue is trying to solve a problem that has and is already been solved.
For V2V, one simply provides:
V2V Tranmitters::
- transmits a strong, ~450 THz (Terahertz) CW (Continuous Wave) signal simultaneously from 3 transmitters (one at each corner, the other, from the center and high, above the bumper, from each corner towards the rear of each car to indicate deceleration.
- transmits a strong, ~450 THz or 520 THz pulsed CW signal at a 1 to 2 Hz pulse rate on the side towards the rear or front of the car and on the optionally on the side of the car that the car intends to turn.
- transmits a broad-spectrum CW signal with frequencies uniformly distributed between ~400 THz and 600 THz from 2 transmitters to the rear of the car when the car is engaging reverse propulsion.
- transmits a broad-spectrum CW signal with frequencies uniformly distributed between ~400 THz and 600 THz from 2 transmitters to the front of the car at night and optionally less bright during the day from each front corner

V2V Receivers:
Simultaneous Reception of these signals from multiple transmitters can be accomplished with multiple CMOS or CCD optical sensor arrays facing each direction from the vehicle, offering directional and signal detection. This system should handle at least 0.2 degrees of angular resolution

Such a system would provide sufficient bandwidth for simultaneous reception from 100% of the vehicles within view of any car with latency rates of less than 1 nanosecond.
Fortunately, the cost for 100% implementation and passing legislation mandating these V2V transmitters, as described above, throughout the entire automobile fleet, in the developed world, is $0.00. Tesla, and many other cars already have the receivers as well, at least in the forward, most critical, direction.

For those who have not followed my technobabble, here's a summary:
V2V Transmitters:
- 3 Red Brake Lights
- Red or Amber turn signal lights
- White backup lights
- White backup or Daytime Running Lights

V2V Receivers:
Camera to observe the lights.

A 2nd generation system might modulate, with simple OOK (On Off Key), the lights to provide additional important details such as deceleration rate, speed, and other capabilities to be determined.

2, 3, or 4G V2I has been available for decades via cellular systems for non-real-time communications. Nothing more is needed. No 5G, 6G, 7G, all the way up to google-plexG (remember when google was a number, not a verb?) is or will be necessary - and it will work much better.

jordanrichard | January 10, 2019

How would V2V help a FSD Tesla know that the supercharger 50 miles down the road is down for the count........?

This isn't nic-picking. Your phone and car depend on a cell signal to know what is ahead. What we are going through now and for the next several years is akin to the transition from there being no "horseless carriages" on the road to no horses on the road..........

TeslaTap.com | January 10, 2019

@Earl and Nagin - Smart!

Earl and Nagin ... | January 10, 2019

@jr,
knowing charging infrastructure status sounds like a V2I problem which would most likely be best solved by a V2I solution, not a quick, real-time V2V solution.

jordanrichard | January 10, 2019

I assume V2I means vehicle to Internet, but that is the rub. That as I pointed out means you have to have an uninterrupted internet connection being it by a cell signal that may or may not be there.

Earl and Nagin ... | January 10, 2019

@jr,
I've assumed "Vehicle to Infrastructure" with the Internet being a logical infrastructure.
I also agree with you that a cell signal is not very reliable, therefore, real-time, safety critical information must not depend on it. I, for one, don't see that "the supercharger 50 miles down the road is down for the count" is real-time or safety critical information although it could cause you to take plan B when you get to the down supercharger. Let's assume that all Superchargers are within cellular coverage so you can at least call Tesla for a tow.
Ideally, a V2I system would get an update on the operational status of all Superchargers within a couple of car ranges while at a Supercharger to ensure one is aware that the next one is down. That would handle all problems except for a Supercharger going down while you were in transit without a cell signal.
Having spent much of my career in the cellphone industry, I would never rely on them for anything critical.

jordanrichard | January 10, 2019

Earl an Nagin, “.....it could cause you to take plan B”. The subject at hand is FSD, where there is no “you”.

It all boils down to is where does the real time data come from, that would never go down and has 100% coverage throughout the U.S.

Earl and Nagin ... | January 11, 2019

@jr,
I'm not sure where you're going: I fail to see why charging station availability is a real-time critical issue. Am I missing something?
The "you", if a responsible person is not in the car, is the car. Since today's Supercharger stations require manually plugging in, I assume there is a person involved at the SC. If the SC is down and there isn't cell coverage, WiFi, etc at the SC, then, clearly, the car isn't going further. I assume it would be missed by someone, eventually, and a physical search party would go out. Presumably, if there was V2I at the SC you left, the infrastructure would have known the destination and there most likely would have been enroute position and direction reports on the way to the down SC, cell, and WiFi so the search would be quite trivial.

jordanrichard | January 11, 2019

I didn't mean ti imply that the supercharger was critical. It is just an example of the type of data that our cars receive via an external source. Though if a FSD car is in need of a charge, how will it know if any given charger is available? This of course means the chargers have the "snake" to self charge a car, but the point is still valid. Any data or rather function in a car, even the basic radio, that is dependent on external sources requires a fool proof source.

As I mentioned earlier, there is one particular area of North west CT that like clock work, my car loses it's cell signal and I can damn near tell you which spot on the road it goes out and 10 mins down the road where it comes back. So, how would a FSD car navigate, using real time traffic information if it can't get that information.?

Earl and Nagin ... | January 11, 2019

@jr,
Ok, we're not too far off the same page. There definitely are gaps in cellular coverage. Any FSD system that relies on real-time cellular connectivity would be a total fail.
I think you're missing what parts need to to totally fool proof and which ones for which the communications is just a convenience.
Personally, I wouldn't trust any FSD system that relies on any kind of cooperative communication with other vehicles or infrastructure to keep out of hazardous situation like running into another car. These should be stand alone sensors such as cameras, radar, and possibly lidar.
Why should FSD count on more than I do as a driver? I don't count on other drivers signalling their turns. I watch for deceleration instead of counting on brake lights. I watch for motion instead of counting on backup lights. FSD should too.

TeslaTap.com | January 11, 2019

I don't think any automaker is looking for V2V or V2I to be present for FSD to work. Clearly if V2V communication was available in every car today, it may take 20-30 years to get 95% coverage.

I look at V2V as a set of incomplete "bonus" information. It allows FSD to get a better picture of the surroundings and may allow it to make better choices.

Interestingly, a V2P (Vehicle to person) has been spec'ed out, and is in the planning states to be put into future phones, so pedestrians with phones would be better identified to FSD cars. Considering how many people are walking around with their head buried in the phone with no clue what's going on around them, it makes a lot of sense.

TeslaTap.com | January 11, 2019

oops: .... in every NEW car sold today....

jordanrichard | January 12, 2019

Earl and Nagin, right, so I will get us closer :-).

Right now, our nav system takes into account traffic condtions to establish a route. Not within sight traffic conditions, but between you and your destination. So, a FSD car would need to utilize the same data to navigate around a traffic jam say 5 miles away. Presently our cars get that from a third party via a cell signal.

Those items you listed that should be stand alone, can’t determine if there is a traffic jam a town or two away. Those items can’t, as I mentioned before, alert the car about a supercharger that may be down and to head to another one.

Basically think of everything we presently have in our cars that gives us a heads up of what is ahead and where does that information come from......

Earl and Nagin ... | January 12, 2019

@jr,
I think you're confusing the NEEDs of FSD with the "Nice-to-haves" of FSD.
IMHO, the NEEDs are to get to the destination without killing or damaging any people or property.
Traffic jam circumnavigation and rerouting around down Superchargers are 'Nice-to-have'. They also can tolerate seconds to minutes of delay.

jordanrichard | January 13, 2019

If the car needs a charge, locating a working charger is not a “nice to have”. Also IMHO, this is not me picking out one outlier situation to make my point. The source of the data to let our cars know the status of super chargers is the same as traffic and that is an outside of the car source.