Forums

Teleoperations for full autonomy

Teleoperations for full autonomy

Does Tesla consider teleoperations to be a step in achieving full autonomy?

ayanmullick | 5 mars 2019

ok

Earl and Nagin ... | 6 mars 2019

Wireless networks probably aren't reliable enough and have too high latency for reasonable teleoperations without a lot of autonomy in the vehicle. I doubt it would be much of a help.
Musk did mentioning being able to steer with your phone while summoning though.

El Mirio | 6 mars 2019

I suspect Teleoperation's to be considered for Tesla Semi, if drones can be steered remotely why not trucks. Not sure whether regulations actually allow to do this tho.

bp | 7 mars 2019

The only teleoperation that might be supported could be with the Semi, which they indicated could be driven in packs of 3 with a driver only in the front Semi. And that would likely require vehicle-to-vehicle communication, since you can't rely on the cell phone network - especially when controlling a fully loaded semi operating at highway speeds...

Flying drones remotely is different than driving a vehicle in heavy high speed traffic, where the reaction time must be much faster.

What's more likely is FSD with remote control of routing and destinations, but not in actual driving of the vehicles. That's what Tesla was planning to do with the Tesla Network (which has now evidently disappeared from the marketing) and what Uber plans to do when they replace human drivers as soon as they can...

bp | 7 mars 2019

The only teleoperation that might be supported could be with the Semi, which they indicated could be driven in packs of 3 with a driver only in the front Semi. And that would likely require vehicle-to-vehicle communication, since you can't rely on the cell phone network - especially when controlling a fully loaded semi operating at highway speeds...

Flying drones remotely is different than driving a vehicle in heavy high speed traffic, where the reaction time must be much faster.

What's more likely is FSD with remote control of routing and destinations, but not in actual driving of the vehicles. That's what Tesla was planning to do with the Tesla Network (which has now evidently disappeared from the marketing) and what Uber plans to do when they replace human drivers as soon as they can...

blue adept | 7 mars 2019

Short answer?

No.

Remotely controlled vehicles are not the path to full autonomy.

ayanmullick | 8 mars 2019

Tesla wants to accelerate the advent of smarter and sustainable transport. I think completely ignoring the significance of teleeoperations might delay FSD in the long run.

https://www.wired.com/story/phantom-teleops/

Earl and Nagin ... | 8 mars 2019

Teleoperated unmanned systems tend to substitute communications bandwidth (mainly RF) for autonomy. Fully teleoperated ground vehicles take a whole lot of very reliable, low latency bandwidth. I'm not saying it can't be done since I've done it but it does not scale to wide consumer usage levels very easily.
Also, since teleoperation relies on a man-in-the-loop, then someone must be paid to watch over it. Therefore, it isn't particularly beneficial, even if bandwidth were cheap, except, perhaps for military drones or Mars mission which fly in harsh areas where it is risky or difficult to put people.

blue adept | 12 mars 2019

That's what I was getting at, E and N, there are some 325.7 million people in the U.S. alone and of those some 76% of them drive!

Just imagine the size of the workforce that would be required to 'monitor' some 247,532 motorists/commuters daily!!

Add to that your mention of the required bandwidth allocation and the whole idea of "teleoperated"/remotely controlled vehicles becomes logistically untenable, and don't even get me stated on liability issues in the instance of an accident!

ayanmullick | 16 mars 2019

In reference to meeting regulations for FSD in residential areas; Waymo wouldn't have been allowed to operate driverless without a teleoperator. Adopting that model of operations could help meet regulations sooner than later.

Earl and Nagin ... | 17 mars 2019

@blue adept,
While it might make regulators happy, it doesn't make sense business wise if you have to have someone sit somewhere driving it anyway. You might has well have the driver in the car -- its a lot cheaper because we know how to do it.

bp | 18 mars 2019

Another reason why Tesla will never implement teleoperations - liability.

While the focus on FSD approval has been on the technology and getting government acceptance, noone's talking about the liability issue - what will the insurance companies do about FSD? Who is responsible when there's an accident? Is it the vehicle owner - or is it the manufacturer, who is essentially providing the driver.

Tesla and the other manufacturers are likely assuming the vehicle owners will be responsible for any accidents, even if the vehicle is operating entirely under FSD without anyone in the vehicle, otherwise the potential downstream cost for the FSD could be huge for the manufacturers, even if the software operates safer than a human driver (accidents will still happen).

Teleoperations clearly would put the responsibility back to Tesla. If a vehicle they were remotely driving gets in an accident - Tesla would likely have complete responsibility for the accident - something they (and the other manufacturers) will likely always try to avoid.

Plus, we see enough problems with the internet today from the central console, with areas of weak or no signals, that teleoperations would be risky.

Probably - not - going - to - happen...

TeslaTap.com | 18 mars 2019

Insurance is an interesting aspect of FSD. If it proves far lower risks, then the insurance company benefits. I think Volvo has said they plan to include insurance with their future FSD solution in Europe. This may run into regulatory issues here in states, so that may not work (i.e. the insurance industry is highly regulated).

As to what happens when FSD makes an error, as far as payments - your insurance company would pay out. Rates may reflect the lower risk too, but I suspect until lower accident rates are proven, we'll be paying the same rate as if driving ourselves.

Now for criminal situations, it's a different matter. If FSD kills a pedestrian, it would seem the liability goes to the car maker.

blue adept | 18 mars 2019

EDIT:

"stated" = started

El Mirio | 18 mars 2019

@TeslaTap Insurance will be interesting indeed. I think a loop whole might be for OEM´s to act as re-insurers offering better premiums to insurance companies, this would avoid OEM´s to deal with the end user insurance intricacies.

blue adept | 18 mars 2019

@ayanmullick

Perhaps, but then Waymo isn't part of the 247,532 thousand individual drivers on the road daily, instead, they are a chauffeured transportation service, in otherwords, a taxi which, last I heard, had limited operations (some 600 self-driving Chrysler Pacifica minivans) in Phoenix, Arizona.

Do not misunderstand me, it (Waymo) is a practical, necessary, even awesome idea/service.

But there's still quite a lot of real world telemetry required for a service that, thus far, has operated only within a "controlled" environment before it can truly be seen as a viable commuter alternative for integration into the state/nation as there are simply far too many unforeseeable environmental variables, and let's not forget what I said about the workforce commitment required to even make the service possible.

Imagine row upon row of cubical after cubical of VR goggle wearing/screen watching office "drone"....

blue adept | 18 mars 2019

@Earl and Nagin ...

I agree completely, though I think that the ultimate goal of services like Waymo is to, eventually, transition to fully autonomous, AI controlled/operated, "hive-minded" vehicles (and now those rows upon rows of manned, VR goggle wearing/screen watching office drones have been replaced by a server farm...'Progress').

blue adept | 18 mars 2019

@bp

Agree, not for a while yet anyway as there'll necessarily need be a rather large infrastructure overhaul/investment to make the potential liability more...diminished...as there is far more telemetry required to make such operations feasible on even a modest scale, definitely far more than what a single vehicle is capable of gathering, let alone processing, in the milliseconds to microseconds to nanoseconds to even picoseconds, required to avoid a potential collision/accident.

There are simply far too many unpredictable variables in both Nature and human behavior requiring computational ability we've not yet managed to design/build/achieve...Believe me, 'does not compute' would not provide sufficient explanation or emotional relief to quell the torture of a child's/loved one's loss in this 'pass-the-blame', scapegoating culture we live in.

blue adept | 18 mars 2019

@El Mirio

So, like the dealerships that offer their own "financing" service for vehicle purchases?

That could pose legal considerations all on it's own inasmuch as the "OEM" would have a 'vested interest' in dispelling any potential claims to ensure the perceived integrity/safety of their FSD service so as to not lower sales (kinda like the quandary Boeing is finding itself in lately).

El Mirio | 18 mars 2019

@blue adept, Re-Insurers, something to learn about. Its a backup insurer, not the same thing as financing a car first hand or be involved in the initial transaction.

blue adept | 18 mars 2019

@ayanmullick

And back to your original question, the one thing that conceptionists/designers/regulators seem to underestimate/overlook is that the only thing consistent about the world is that it is not static, it is dynamic and ever changing...Hard as hell, nigh impossible, to compensate for the inherent fluidity of Nature, human or otherwise.

El Mirio | 18 mars 2019

@blue adept as fast as possible "back to your original question"

ayanmullick | 30 mars 2019

@bp, Liability avoidance, in my opinion, isn't the best strategy. Liability will have to be owned by the manufacturer. We have to be far along in the march of 9's to ensure there are no fatalities.

@El Mirio, Insurance is statistics-based. Once we have data that a specific stretch of road is safer for FSD than manual-driving, insurance will probably take care of itself.

@blue adept, Conversion of blue collar jobs to white collar jobs is a force of nature. This is inevitable.

ayanmullick | 30 mars 2019

Liability avoidance, in my opinion, isn't the best strategy with respect to accelerating the advent of FSD.*

March of 9's: https://www.youtube.com/embed/MOy3MTZ9Dyg?start=600&end=640