Autopilot lane keeping still not available over 6 months after delivery

[MsElectric]

I wonder if the next generation Autopilot Sensor Suite will be based on NVidia PX technology and sensors INSTEAD of Mobile Eye or it will be a combination of Mobile Eye and NVidia PX.

I also wonder what degree of Autopilot capabilities we will actually see implemented without rear and side radar (both present in the Mercedes implemntation).

 

[LetsGoFast]

I thought the D and Autopilot reveal was October 9, 2015 (not September).

Pretty sure it was in the past. 2014 perhaps. :tongue:

 

[anxman]

I ordered my P85D the evening that it was unveiled (technically I ordered it a week earlier but my rep couldn't confirm its existence).

My understanding at the time of delivery was that the autopilot software was going to be available sometime in 2015. My expectation was sometime around November or December of 2015. There was a WSJ article around December that Tesla was amazing to have it available by H1 2015 but I understood that to mean more like August.

 

[Soolim]

I wonder if the next generation Autopilot Sensor Suite will be based on NVidia PX technology and sensors INSTEAD of Mobile Eye or it will be a combination of Mobile Eye and NVidia PX.

I also wonder what degree of Autopilot capabilities we will actually see implemented without rear and side radar (both present in the Mercedes implemntation).

So if in the near future (say next month or two) Tesla give us the autopilot that is less than what Mercedes and BMW are capable of, due to the inadequacy of the existing sensor, would Tesla be proud to announce that? This is not to rule out that Tesla can achieve the same performance without the additional sensors since I am not educated enough. But I doubt it.

I am speculating that Tesla is rethinking the sensor and autopilot package of MS, and is actively testing an improved system that would at least match Mercedes/BMW in the MX, hence the silence and the delay of autopilot in MS.

 

[Stoneymonster]

So if in the near future (say next month or two) Tesla give us the autopilot that is less than what Mercedes and BMW are capable of, due to the inadequacy of the existing sensor, would Tesla be proud to announce that? This is not to rule out that Tesla can achieve the same performance without the additional sensors since I am not educated enough. But I doubt it.

I am speculating that Tesla is rethinking the sensor and autopilot package of MS, and is actively testing an improved system that would at least match Mercedes/BMW in the MX, hence the silence and the delay of autopilot in MS.

Nope. They are behind on the software they have promised for the existing sensor suite, which is sufficient for the currently promised feature set. Nothing more, nothing less.

Any advancements are for model updates later on, they are not "holding back" because they figured out they couldn't do it. If that is true, I won't buy another Tesla, because that means they are absolutely terrible engineers.

 

[green1]

the existing sensor suite, which is sufficient for the currently promised feature set. Nothing more, nothing less.

sufficient for everything Tesla has promised, however for some reason a lot of people heard all sorts of things that tesla never said.

 

[Soolim]

Quote from the Tesla Website:
Autopilot combines a forward looking camera, radar, and 360 degree sonar sensors with real time traffic updates to automatically drive Model S on the open road and in dense stop and go traffic. Changing lanes becomes as simple as a tap of the turn signal. When you arrive at your destination, Model S will both detect a parking spot and automatically park itself. Standard equipment safety features are constantly monitoring stop signs, traffic signals and pedestrians, as well as for unintentional lane changes.

This is what is promised for NOW. I don't think they are having trouble with the software ONLY. How would the current sensors be adequate for dense stop and go traffic (i.e. city driving)? Software can only do so much if the measured detected range are adequate for response time. Somewhere I read that the sonar installed has a range of 16 ft. Is 16 ft detection range adequate?
Current sensors are already inadequate as park assist (for coil suspension) in area with low curb that is not detectable by sonar or radar. How would the self parking be functional with current sensor deficiency? I speculate that some additional cameras or sensors would be required even if just for self parking.

I hope my lack of optimism is wrong. Time will tell.

 

[Stoneymonster]

Quote from the Tesla Website:
Autopilot combines a forward looking camera, radar, and 360 degree sonar sensors with real time traffic updates to automatically drive Model S on the open road and in dense stop and go traffic. Changing lanes becomes as simple as a tap of the turn signal. When you arrive at your destination, Model S will both detect a parking spot and automatically park itself. Standard equipment safety features are constantly monitoring stop signs, traffic signals and pedestrians, as well as for unintentional lane changes.

This is what is promised for NOW. I don't think they are having trouble with the software ONLY. How would the current sensors be adequate for dense stop and go traffic (i.e. city driving)? Software can only do so much if the measured detected range are adequate for response time. Somewhere I read that the sonar installed has a range of 16 ft. Is 16 ft detection range adequate?\
Current sensors are already inadequate as park assist (for coil suspension) in area with low curb that is not detectable by sonar or radar. How would the self parking be functional with current sensor deficiency? I speculate that some additional cameras or sensors would be required even if just for self parking.

I hope my lack of optimism is wrong. Time will tell.

Because dense stop and go traffic is primarily forward looking (the radar can see about 150-200ft.). Lane changes are initiated by the driver, and the side looking ultrasonics are sufficient for that (plus a mirror check). They committed to self parking with the current sensors, why do you think a random internet forum knows more about the details than the engineering team?

The second guessing here is silly.

 

[MarcG]

Quote from the Tesla Website:
Autopilot combines a forward looking camera, radar, and 360 degree sonar sensors with real time traffic updates to automatically drive Model S on the open road and in dense stop and go traffic. Changing lanes becomes as simple as a tap of the turn signal. When you arrive at your destination, Model S will both detect a parking spot and automatically park itself. Standard equipment safety features are constantly monitoring stop signs, traffic signals and pedestrians, as well as for unintentional lane changes.

This is what is promised for NOW. I don't think they are having trouble with the software ONLY. How would the current sensors be adequate for dense stop and go traffic (i.e. city driving)? Software can only do so much if the measured detected range are adequate for response time. Somewhere I read that the sonar installed has a range of 16 ft. Is 16 ft detection range adequate?
Current sensors are already inadequate as park assist (for coil suspension) in area with low curb that is not detectable by sonar or radar. How would the self parking be functional with current sensor deficiency? I speculate that some additional cameras or sensors would be required even if just for self parking.

I hope my lack of optimism is wrong. Time will tell.

I believe the "stop and go traffic" mention is meant for congested highways, not city driving. If this is the case, the current hardware sensors should be enough, as all the car needs to do is stay in its lane and follow the car ahead (the latter of which is already accomplished with TACC).

As for self-parking, I think the 16-foot range of the ultrasonic sensors is also enough when combined with the rear camera. I personally would not want the car to self-park, as I have not seen a single proper implementation of self-park feature by any manufacturer. Having grown up driving in Europe and learned to park in the tiniest places, I don't think a car could do a better job... but maybe I'm being too cocky on this one :wink:

 

[Todd Burch]

This is what is promised for NOW. I don't think they are having trouble with the software ONLY. How would the current sensors be adequate for dense stop and go traffic (i.e. city driving)? Software can only do so much if the measured detected range are adequate for response time. Somewhere I read that the sonar installed has a range of 16 ft. Is 16 ft detection range adequate?
Current sensors are already inadequate as park assist (for coil suspension) in area with low curb that is not detectable by sonar or radar. How would the self parking be functional with current sensor deficiency? I speculate that some additional cameras or sensors would be required even if just for self parking.

I hope my lack of optimism is wrong. Time will tell.

Yes, you're wrong. There is no "sensor deficiency". You will find plenty of cars that can self park and do a lot of this stuff with similar hardware.

Do you really think Tesla would have sold tens of thousands of cars with a hardware package that they weren't sure would be adequate?

I'll answer a few of your questions.

Current sensors are NOT inadequate for park assist. The further the car is from the curb, the greater the field of view that the sensors have. The sonar system can detect a curb from a few feet away, and spatially track the position of the car to that curb. It's not rocket science. Not trivial, but it's been done many times before.

Science experiment: Curl your hands around your eyes like you're looking through a pair of binoculars. What you can see is now what the sonar sees. Now, look straight across the room toward something that you can see on the floor (a table, for instance). Walk toward it, holding your "binoculars" level. Walk right up to it so that it's near your feet but so you can't see it anymore. You should be able to do this easily, even though you can't see the table anymore, because your brain remembered the spatial position of the table or chair. The sonar can do this too, only more accurately.

16 feet is plenty far enough for self parking. It's not like the car will be speeding at 35 mph into a parking spot.

And for lane changes, don't assume that the car will check to ensure that a lane change is safe for you. Tesla never promised that. It can ensure your blind spot is clear and that there's no vehicle next to the car, but the driver will be responsible for ensuring that a speeding car in the left lane isn't coming up, making a lane change unsafe.

 

[Soolim]

Current sensors are NOT inadequate for park assist. The further the car is from the curb, the greater the field of view that the sensors have. The sonar system can detect a curb from a few feet away, and spatially track the position of the car to that curb. It's not rocket science. Not trivial, but it's been done many times before.

You mean to tell me that the front radar/sensor will remember the height of a curb stop in front of the car, and one can rely on that to park without scraping the bottom of the bumper using park assist? I have just been to a service center and they told me that some people do scrape the front bumper on low curb (but still higher than bumper clearance). This is of course assuming that the driver is careful, but information could still be missing on the monitor/screen.

Many cars do not have the low clearance such as that of MS with coil suspension. Hence it is not a problem for those cars.

 

[MileHighMotoring]

I believe the "stop and go traffic" mention is meant for congested highways, not city driving.

I have to say - I think the opposite. I don't need city self-driving, but "Stop And Go", to me, is the euphemistic phrase meaning the opposite of Highway Driving.

 

[Todd Burch]

You mean to tell me that the front radar/sensor will remember the height of a curb stop in front of the car, and one can rely on that to park without scraping the bottom of the bumper using park assist? I have just been to a service center and they told me that some people do scrape the front bumper on low curb (but still higher than bumper clearance). This is of course assuming that the driver is careful, but information could still be missing on the monitor/screen.

Many cars do not have the low clearance such as that of MS with coil suspension. Hence it is not a problem for those cars.

I don't know what it *will* do (only Tesla knows that). I can only tell you what it *can* do. And yes, the sensors/ hardware can do that. How do you explain all the other cars with identical or less hardware that are on the market that can do it?

And the fact that other self-parking cars don't drive over parking barriers is not because they have a few more inches of ground clearance.

- - - Updated - - -

OK, I put a YouTube video where my mouth is:

What is used for auto-parking this 2012 Ford Focus? Ultrasonic sensors. (Looks like there are less than the Model S, actually).

I don't see the Focus driving over any curbs.

 

[Soolim]

As for self-parking, I think the 16-foot range of the ultrasonic sensors is also enough when combined with the rear camera. I personally would not want the car to self-park, as I have not seen a single proper implementation of self-park feature by any manufacturer. Having grown up driving in Europe and learned to park in the tiniest places, I don't think a car could do a better job... but maybe I'm being too cocky on this one :wink:

I am in the minority here. I want the self parking more than the auto lane change. In tight spot, I would like to be out of the car and let the car park itself, as shown by competitor's video.

 

[MileHighMotoring]

I am in the minority here. I want the self parking more than the auto lane change. In tight spot, I would like to be out of the car and let the car park itself, as shown by competitor's video.

+1 Cars have been parking themselves for at least 5 years now.

 

[Soolim]

I don't see the Focus driving over any curbs.

Thanks for the video of Ford Focus parallel parking. When I said hitting the curb stop (with front or rear bumper), I am referring to parkade stall parking. I will experiment with 2"x 4" wood studs and see if the front radar/sensor and rear sensor remembers the height of the 2x4 wood stud without passing it under the bumper to have a peace of mind.:smile:

 

[AmpedRealtor]

I wonder if the next generation Autopilot Sensor Suite will be based on NVidia PX technology and sensors INSTEAD of Mobile Eye or it will be a combination of Mobile Eye and NVidia PX.

I also wonder what degree of Autopilot capabilities we will actually see implemented without rear and side radar (both present in the Mercedes implemntation).

Bet you we see this in Model X.

 

[Soolim]

If Tesla give us the self-parking in July 2015, I would be happy. Since Tesla do incremental upgrades on software, and if the hardware is in fact adequate and ready, my request would be for Tesla to release the self park feature ASAP. Like some of you said, it is a 4-5 years old technology, so why can't we have it today considering that some of us has prepaid for the package?

 

[Todd Burch]

Thanks for the video of Ford Focus parallel parking. When I said hitting the curb stop (with front or rear bumper), I am referring to parkade stall parking. I will experiment with 2"x 4" wood studs and see if the front radar/sensor and rear sensor remembers the height of the 2x4 wood stud without passing it under the bumper to have a peace of mind.:smile:

Yes, I am aware that you're talking about parkade style parking. My point was that the same idea applies.

Be careful though. Don't confuse auto-parking with the existing parking sensors. I'm talking about two different things here. The existing parking sensors as they're implemented won't necessarily go off if the curb stop is out of view of the sensors.

- - - Updated - - -

Like some of you said, it is a 4-5 years old technology, so why can't we have it today considering that some of us has prepaid for the package?

Because Tesla has been around a total of 10 years or so, most of which they've spent developing an entirely new drivetrain technology. Other car companies have been around for 100+ years. Tesla has to play catch-up, since you can't exactly buy auto-parking systems off the shelf and glue them onto a car.

 

[Max*]

Thanks for the video of Ford Focus parallel parking. When I said hitting the curb stop (with front or rear bumper), I am referring to parkade stall parking. I will experiment with 2"x 4" wood studs and see if the front radar/sensor and rear sensor remembers the height of the 2x4 wood stud without passing it under the bumper to have a peace of mind.:smile:

It can do it. I posted in another thread (mobile now, or I'd link it), I have a bumper block about 4-5" of the ground. Tesla sees it, tells me how far it is, then it disappears and it measures the distance to the next wall (let's say 18"), as I'm rolling forward it yells STOP. I'm no where near the wall, but it remembered the location of the bumper block.

If it can see it now. And track it. It can park it when the software is released.