Tam
Well-Known Member
Tesla promised HW1 to be able to do hands-free On-ramp to Off-ramp.
Owners paid for that promise even before that task could be fulfilled. Now, it is time to fulfill it. That's enough rationale.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla promised HW1 to be able to do hands-free On-ramp to Off-ramp.
Owners paid for that promise even before that task could be fulfilled. Now, it is time to fulfill it. That's enough rationale.
I'm beginning to question that myself.
I informed a trucker with about 30 years experience how the additional cameras would enable full autonomy on new Teslas. I was then informed that radar has proven to be superior to cameras on their trucks which have been operating like AP1 on Teslas for many years. They tried using cameras on the trucks but removed them because they didn't work as well as the radar. Since Tesla has radar, this trucker didn't understand why Tesla even needed the additional cameras except to give the appearance to automobile novices of improved safety.
Maybe a trucker will chime in one way or the other.
Tam
Well-Known Member
The demonstration of Otto truck that hauled 50,000 cans of beer without a driver at the wheel on 120 mile freeway stretch was equipped with:
3 LIDARs
1 RADAR
1 Camera
It's a sensor fusion from different kinds that gives the system a complete picture/data.
It is a mistake to drop a sensor if you can add another kind of sensor such as radar vs camera.
Last edited:
for AP1 cars, I would like to see:
* an option for a fee, to replace the older sonars with newer sonars. I am assuming this is a 1 for 1 replacement to give better & faster feed to AP1 system. Current blind spot detection is non-functional at freeway speeds, hopefully new sonars will improve things.
* remove the gimmicky auto lane change feature and add traffic signal reading. If the code is similar, it will benefit both AP1 & AP2 cars.
* if there is any room remaining on current CPU, add the same code that AP2's rear camera feed processing uses, into AP1 system. This will make the AP1 systems utilize 2 cameras and the radar, and should make for a safer and better autopilot system.
I can totally see why Tesla may not want to do any of the above, particularly if, the s/w dev teams for AP1 and AP2 are separate and there is no leverage. But they are smart folks, and leveraging code across both platforms would be something they would have already planned. e.g. AP1 uses 1 camera+ 1 Radar, and all that s/w code is reused into AP2 systems + feed from additional cameras used for better real-time decision making and additional features. I am quite hopeful this is the case (leveraging code across both platforms) and thus, AP1 cars would continue to improve, however few those enhancements might be.
* an option for a fee, to replace the older sonars with newer sonars. I am assuming this is a 1 for 1 replacement to give better & faster feed to AP1 system. Current blind spot detection is non-functional at freeway speeds, hopefully new sonars will improve things.
* remove the gimmicky auto lane change feature and add traffic signal reading. If the code is similar, it will benefit both AP1 & AP2 cars.
* if there is any room remaining on current CPU, add the same code that AP2's rear camera feed processing uses, into AP1 system. This will make the AP1 systems utilize 2 cameras and the radar, and should make for a safer and better autopilot system.
I can totally see why Tesla may not want to do any of the above, particularly if, the s/w dev teams for AP1 and AP2 are separate and there is no leverage. But they are smart folks, and leveraging code across both platforms would be something they would have already planned. e.g. AP1 uses 1 camera+ 1 Radar, and all that s/w code is reused into AP2 systems + feed from additional cameras used for better real-time decision making and additional features. I am quite hopeful this is the case (leveraging code across both platforms) and thus, AP1 cars would continue to improve, however few those enhancements might be.
Once the Model 3s are produced, AP 1.0 will be quickly dwarfed by AP 2.0.
Unless someone can confirm AP 1.0 uses the same "source code" as AP 2.0, we can assume the software for AP 1.0 is different than 2.0 - and what Tesla likely did was "port" the autopilot software to the new AP 2.0 language/code environment - and if that was done, it means additional coding will be required to add any changes into the AP 1.0 code.
Coupled with the hardware limitations, AP 1.0 should be realistic about expectations for future improvements. Tesla plans to enable on ramp to off ramp functionality - and AP 1.0 should continue to see improvements because of the improving quality of the data driving autopilot (which should be shared between 1.0 & 2.0).
But additional functionality beyond what is planned for 8.1 seems less likely.
And liability is a major risk for future limitations on AP 1.0 (and 2.0) functionality. So far, Tesla has been able to avoid (as far as we know) any liability for accidents while AP was in operation. It's inevitable that Tesla will eventually be held liable for an AP-related accident, and when that happens, Tesla may be forced or voluntarily impose more restrictions on all AP usage, such as geolocation limits on what AP 1.0 can be activated (such as only on limited access highways).
Unless someone can confirm AP 1.0 uses the same "source code" as AP 2.0, we can assume the software for AP 1.0 is different than 2.0 - and what Tesla likely did was "port" the autopilot software to the new AP 2.0 language/code environment - and if that was done, it means additional coding will be required to add any changes into the AP 1.0 code.
Coupled with the hardware limitations, AP 1.0 should be realistic about expectations for future improvements. Tesla plans to enable on ramp to off ramp functionality - and AP 1.0 should continue to see improvements because of the improving quality of the data driving autopilot (which should be shared between 1.0 & 2.0).
But additional functionality beyond what is planned for 8.1 seems less likely.
And liability is a major risk for future limitations on AP 1.0 (and 2.0) functionality. So far, Tesla has been able to avoid (as far as we know) any liability for accidents while AP was in operation. It's inevitable that Tesla will eventually be held liable for an AP-related accident, and when that happens, Tesla may be forced or voluntarily impose more restrictions on all AP usage, such as geolocation limits on what AP 1.0 can be activated (such as only on limited access highways).
The next goal for AP2 is parity with AP1 functionality. Why in the world would they increase AP1 functionality before that, to set the goal higher? If anything, they'll decrease AP1 functionality...
I wouldn't put much stock in anything said by an end user of such systems to date. Recent advances in machine learning have made camera-based systems(not just for cars) far more capable than they used to be. I find it unlikely that any trucking company had the resources a few years ago to find sufficient data to train and develop a network based on their cameras.
Radar is probably better using very basic, straightforward algorithms for object detection and the like. But add a good neural network implementation on top of that and the whole landscape changes.
Tam
Well-Known Member
Safety is an incentive.
AP1 has been here first so it should get improvement first. The faster you make sure that AP1 can reliably stop for a stationary vechicle, the better.
The faster you can make sure AP1 won't hit any guardrails when cornering, the better.
AP2 comes later but it will catch up much faster and will become much better by the end of this year with demonstration of driverless cross-country trip due to its Artificial Intelligence, processing power and better sensors.
Last edited:
Isn't AP1 largely MobileEye, with Tesla's software running on top? In which case, given the relationship between the two companies, I wouldn't expect a lot of additional improvements.
Reading stop signs and traffic signals would be great, and fully within MobileEye's capabilities, but I don't know if the two companies will work together to add them. Same for pedestrian detection.
Reading stop signs and traffic signals would be great, and fully within MobileEye's capabilities, but I don't know if the two companies will work together to add them. Same for pedestrian detection.
In heavy rain this weekend in bay area, our front camera's visibility was so impaired that autopilot (autosteer) was disabled by the system. I got message on dashboard that autosteer was not available. Radar based TACC was still working and I was actually very glad that it was - as my visibility was severely impaired and i depended on the radar to do any braking based on traffic in front that I (& the camera) couldn't see. I feel AP2 isn't the final HW config - they need to add additional radars, so AP3 may not be impossible (likely in Model3?).
BTW, was in New Zealand on vacation, and rented a mid-size SUV - Holden (Korean?) which had rear cross-traffic detection (quite likely using angled radars on both ends of the rear bumper) - and it was fantastic. During backing up, the dashboard warned me much ahead of seeing cars on the display, in which direction they were coming. WoW! Miss them on our expensive Teslas.
Even a tiny bit of modification on the hw setup will certainly require realworld test of millions of miles before enabled. Which will prevent use of valuable and limited resources of the company for development of newer and more advanced models. Does this sound logical?
Whats going to happen is ap1 cars will be stamped as level1 or 2 autonomy cars and thats it.
Whats going to happen is ap1 cars will be stamped as level1 or 2 autonomy cars and thats it.
3Victoria
Active Member
I think MBLY already can do sign recognition; but perhaps it is not flexible enough for Tesla.
I'm still waiting for summon 2.0 for ap1 that promised our cars would exit the garage and come pick us up at our front door(staying of private property of course).
They of course eventually removed this from their web site but not before the waybackmachine archived it.
RogerHScott
Active Member
"Support" and "continued advancement" are very different things. Tell me about the cell phone of yours that continued to improve (the
phone itself, not the service) years after you bough it.
Anything that slows down their development of HW2 support (like, say, further HW1 advancements) will hurt the overall company far more in the long run than some vague sense of lost entitlement among the aging HW1 population (who, which each passing month, are
further self identifying as "economically uninteresting" to Tesla because they don't buy something new).
RogerHScott
Active Member
There's no way the HW1 sensor suite will ever be able to do this reliably enough to be "safe", as in "taking over any safety burden
from the driver". If, heaven forbid, the law ever requires that capability for the other AP features to be considered "safe" Tesla
would have no alternative but to disable AP1 entirely.
Every smartphone I've ever owned. They've always received new OS' with new features at least a couple years after my buying them. And that's despite resale value not being something smartphone manufacturers really need to worry about. Where for cars, if the resale value collapses dramatically, it's going to massively hurt future sales.
There's no direct gain for Tesla by supporting older Model S' but there's plenty of indirect gain from doing so, particularly given the degree to which Tesla relies on brand name.
It can do it very safely. When it reads a stop sign or red light, it stops and goes into "hold" mode. When the driver decides it is safe to resume, he or she presses the gas, and the car resumes travel. Perfectly safe, and perfectly do-able with the AP1 sensor suite.
Hugh Mannity
Mediocre Member
habanero69
I Dont Need Cialis. I Drive an EV.
Wrong on lights and signs. How is a black and white camera going to know the color of a light? That is not going to happen and NOT required now nor likely in the future. AP1 will NEVER be autonomous. It is AutoPilot only...
Similar threads
