-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Can level 5 Autonomy be achieved with Hardware suite 2.0?
- Thread starter Øivind Hoem
- Start date
JeffK
Well-Known Member
Yeah, you're right it looks like the wide angle radar (which would detect in other lanes) only goes 60 meters
Tam
Well-Known Member
I understand the new description as: Driver does not have to "engage the turn signal" to activate Auto Lane Changing.
The autopilot page further clarifies that:
"Once on the freeway, your Tesla will determine which lane you need to be in and when."
I understood that as: It is no longer the driver who needs to "Perform visual checks to make sure it is safe and appropriate to move into the target lane" and "which lane" left or right--The EAP takes over all the decisions for Auto Lane Changing.
Although the automation does not require human input such as steering, braking, decelerating however, the overall principle of Autopilot and Enhance Autopilot is: Human is responsible. You need to watch it and intervene appropriately at ALL TIMES.
If you do not want to be responsible, then get out of the driver seat and pay for Driverless Option and let the car drives itself.
According to definition, Level 5 handles all driving conditions at least as good as average human. So it doesn't need to handle situations that average human can't handle.
"the full-time performance by an automated driving system of all aspects of the dynamic driving task under all roadway and environmental conditions that can be managed by a human driver"
Source http://www.sae.org/misc/pdfs/automated_driving.pdf
Edit; Original source doesn't contain the word "average", but I think it is clear we're talking about average human.
Last edited:
You might want to take a closer look at what they are defining as the "Dynamic Driving Task". They have intentionally separated the "intelligence" required to drive a car into tactical and strategic roles with the latter being outside of the scope of even level 5. This makes perfect sense from a practical standpoint, as AI would take decades to be as good as any human driver in that regard.
JeffK
Well-Known Member
Consider a very simple scenario... I want to go to the supermarket but want to drop something off at the post office on the way.
I hop inside a level 5 car and what happens?
Nothing happens! Why?
JeffK
Well-Known Member
The types of decision making, when confronted with unidentifiable situations, that would be required to make progress. That decision making is not considered part of the level 5 "dynamic driving task". For example, a human directing traffic, parking garage navigation, or entry to roadway requires use of mirror, or listening for approaching traffic around a corner, road conditions such as flooding, fire (heat), tire hazards around construction, road closures, etc. People are good at anticipating and avoiding unfavorable conditions that may be identified a half mile away, or in the city, several blocks away, yet a level 5 car not so much.
The most common problems not handled by a level 5 car would be following the directions signaled/communicated by a human, working out timing at the busy intersection with traffic light out, finding the entrances/exits to various locations from the street, finding the most appropriate place to park at some destination.
Reeler
Decade of Pure EV Driving
With AP 1.0 we were promised blind spot detection, auto parking, freeway exit, etc. and I don't think the sensors can ever do those well enough to be practical. Remember how we were promised infallible speech recognition and it still isn't 100% accurate? Level 5 is not possible in my estimation and we are decades away. Getting a word wrong in speech recognition is not a big deal, but with AP that could be a person being run over. AP will have limited use like cruise control, but we will not be sleeping in the back seat while our car drives anytime soon.
In this video at 8:36 Elon says the following. Here is a direct link to correct moment.
I think what is going to happen is, the Enhanced Autopilot will able to change lanes by itself as long as there is at least one other car behind it on the target lane within 100 meters range that drives around the same speed. For example, let's say the Tesla is driving at 80 mph and another car is 90 meters behind in the left lane, approaching at 84 mph. In this situation, the Tesla would do a lane change by itself without the driver doing anything because the car knows with certainty that the distance is enough. However, in the same situation, if there is no other car behind within 100 meters range, the Tesla won't do a fully automatic lane change because there might be cars more than 100 meters away approaching at much higher speeds.
I think what is going to happen is, the Enhanced Autopilot will able to change lanes by itself as long as there is at least one other car behind it on the target lane within 100 meters range that drives around the same speed. For example, let's say the Tesla is driving at 80 mph and another car is 90 meters behind in the left lane, approaching at 84 mph. In this situation, the Tesla would do a lane change by itself without the driver doing anything because the car knows with certainty that the distance is enough. However, in the same situation, if there is no other car behind within 100 meters range, the Tesla won't do a fully automatic lane change because there might be cars more than 100 meters away approaching at much higher speeds.
Last edited:
JeffK
Well-Known Member
Neural networks cannot be 100% and there's good reason why, the aim is for better than human which is an achievable goal. As far as speech recognition, Microsoft had trained a model last year that was better than a human.
As far as level 4/5 it's definitely possible and sooner than you think.
JeffK
Well-Known Member
On US roads if the Tesla is travelling at 80 mph it's extremely unlikely that a car in the left lane is going so much faster than that in which there's not enough time for the vehicle to react.
Second, I'd argue that side mirrors provide both a limited viewing angle giving you blindspots. As a human you have to take your eyes off the road to both check the mirror and check the blindspot. The majority of humans are also not looking over 100m to the rear in their side mirror. The majority of humans if they see a car 100 meters back will still change lanes without having a perfect notion of how fast the other car is going. In much of the US if you rear end someone then it's your fault.
So, that said, the AP2 computer:
- Has a better viewing angle than a human
- Is watching farther than a typical human
- Is able to more accurately gauge speeds than a human
- Is able to watch the road ahead at the exact same time as registering the views from other angles
- Can react faster than a human
- etc.
If they put a long distance radar able to view 200 meters, then people would still make the same argument... claiming even that is not sufficient when even 100 meters is better than a typical human.
It's not going to be programmed to purposely avoid an edge case where a drunk in a supercar is going 200+ mph in the left lane, at least here in the US.
If I'm going 80mph and a car in the left is going let's say even 110 mph that's a delta of 30 mph (13.41 m/s). You said the car is over 100m away so 100/13.41 = ~7.5 seconds for a reaction by the speeding driver.
one one thousand, two one thousand, three one thousand,... Plenty of time, no worries.
Last edited:
3Victoria
Active Member
The Nvidia set of videos from CES is educational. They say that their hw can use a variety of sensors, including ultrasound, LIDAR, video, speed, GPS, etc., and that different manufacturers will choose their particular suite. I am guessing that Tesla has determined that 8 cameras, sonar and the front radar is sufficient.
Tam
Well-Known Member
JeffK
Well-Known Member
I know you're not going to want to hear this but he did have his blinker on before the truck in front even moved out of the way... If you wouldn't have hit the accelerator so hard you both would have entered into the lane gracefully.
his focus was in his side mirror on the white Ford. An autonomous car would have been able to see all the views at once and might have been able to see you cutting over and react by cancelling the lane change but this driver was already committed and had already checked to make sure the lane was clear (and it was), but he sure wasn't expecting you to dart over and accelerate (in his blind spot no less).
Last edited:
I think this is the best place to put this, not sure.....this is more about competition I guess you could say or you could look at it as another check! that Tesla is far ahead of everyone else:
ARPA-E Funding $32 Million for 10 Reduced-Energy Vehicle Projects - HybridCars.com
Now why would GM need to get 4.2M from the gov't when Tesla developed all of this on their own.....hmmm.......
ARPA-E Funding $32 Million for 10 Reduced-Energy Vehicle Projects - HybridCars.com
Now why would GM need to get 4.2M from the gov't when Tesla developed all of this on their own.....hmmm.......
HookBill
Member
Similar threads
- Article
