Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

What will happen within the next 6 1/2 weeks?

Which new FSD features will be released by end of year and to whom?

  • None - on Jan 1 'later this year' will simply become end of 2020!

    Votes: 106 55.5%
  • One or more major features (stop lights and/or turns) to small number of EAP HW 3.0 vehicles.

    Votes: 55 28.8%
  • One or more major features (stop lights and/or turns) to small number of EAP HW 2.x/3.0 vehicles.

    Votes: 7 3.7%
  • One or more major features (stop lights and/or turns) to all HW 3.0 FSD owners!

    Votes: 8 4.2%
  • One or more major features (stop lights and/or turns) to all FSD owners!

    Votes: 15 7.9%

  • Total voters
    191
This site may earn commission on affiliate links.
Honestly, I don't know for sure. But from what I have seen, Waymo seems to have really good situational awareness of different objects and path prediction. So if Waymo has the driving policy for how to handle this specific driving scenario, then yes, I would think they would be able to handle it.

I've seen this talked about it some presentations. The car must not only predict the actions of other other road users but also predict how those actions will change the actions of other road users. Here is something similar in a Waymo video. The car predicts that a cyclist will go around a truck parked in the bicycle lane.

As much as I love Tesla for the way it's brought EVs into the mainstream, and for building two of the three EVs I've owned, I think Waymo is the company to watch for driverless cars. My dream car would be a Tesla with Waymo FSD technology.

(My first EV was an absolute P.O.S. Zap Xebra. I loved it because it was electric. But it was Tesla that started making proper electric cars.)

In the video, the car predicts that the cyclists will go around a stationary obstacle. I think it increases the difficulty when it's a car that will move aside for a moving cyclist. Note that the car could actually remain in its lane without striking the cyclist. But a thoughtful driver will give the cyclist extra room. In the Waymo video, the cyclists have to go around the parked truck. Also, when approaching a cyclist going the same direction as the car, on the road shoulder, outside the driving lane, will the FSD car move over a bit to give the cyclist extra room? EAP does not. I always disengage to give cyclists and pedestrians extra room even when I would not hit them if I stay centered in the lane. I want my (future) FSD car to do the same.
 
As much as I love Tesla for the way it's brought EVs into the mainstream, and for building two of the three EVs I've owned, I think Waymo is the company to watch for driverless cars. My dream car would be a Tesla with Waymo FSD technology.

Yes, don't let anybody fool you, Waymo has the best autonomous driving tech right now by far.
 
There has actually been a few years of really good research into using neural networks to predict pedestrian paths especially, and some great progress has been made. It's not a solved problem, but I think we're probably closing in on as accurate as a computer will be.

Some links for anybody interested.

Accurate pedestrian path prediction using neural networks - IEEE Conference Publication

StarNet: Pedestrian Trajectory Prediction using Deep Neural...

http://www.gavrila.net/eccv14.pdf

http://cvrr.ucsd.edu/publications/2017/ITSC2017-DeoTrivedi-PedestrianMotion.pdf

http://digitalcommons.calpoly.edu/cgi/viewcontent.cgi?article=2952&context=theses

Pedestrian Path Forecasting in Crowd
 
There has actually been a few years of really good research into using neural networks to predict pedestrian paths especially, and some great progress has been made. It's not a solved problem, but I think we're probably closing in on as accurate as a computer will be.

Some links for anybody interested.

Accurate pedestrian path prediction using neural networks - IEEE Conference Publication

StarNet: Pedestrian Trajectory Prediction using Deep Neural...

http://www.gavrila.net/eccv14.pdf

http://cvrr.ucsd.edu/publications/2017/ITSC2017-DeoTrivedi-PedestrianMotion.pdf

http://digitalcommons.calpoly.edu/cgi/viewcontent.cgi?article=2952&context=theses

Pedestrian Path Forecasting in Crowd

Wonder if that includes drunk pedestrians ? :)
 
  • Funny
Reactions: DrDabbles
HW3 upgrade waiting room: All FSD Tesla’s

Less than 4% of poll respondents have received HW3 update so far!

When real FSD becomes available, or even Level 3 for both city and highway that allows me to take my attention off the road until the car alerts me, I probably won't pay for the FSD upgrade and wait for my turn at the hardware upgrade. I'll just trade in my car. My Model 3 will probably be five or ten years old by then anyway. Assuming I'm still above ground.

Wonder if that includes drunk pedestrians ? :)

We've all had this experience: You're walking along and you approach someone standing still and facing away from you. Suddenly, without warning, they turn around and start walking without looking. Even people cannot predict people. But once a car learns to tell that an object is a person, it will be able to react more quickly than a human driver, and being fully aware of everything around it for 360°, it will be able to choose the action most likely to avoid hitting the person.
 
We've all had this experience: You're walking along and you approach someone standing still and facing away from you. Suddenly, without warning, they turn around and start walking without looking. Even people cannot predict people. But once a car learns to tell that an object is a person, it will be able to react more quickly than a human driver, and being fully aware of everything around it for 360°, it will be able to choose the action most likely to avoid hitting the person.

There was an angry reddit thread about a car that stopped because a pedestrian was looking across the road and wasn't on the sidewalk. Basically every warning sign anybody would look for to expect that person to just walk out into the street. So the car stopped, owner said it "slammed on the brakes", and was annoyed it made the safest autonomous choice possible.

I often want to see peoples driving records when they complain about stuff like that. :D
 
  • Like
Reactions: DanCar
There was an angry reddit thread about a car that stopped because a pedestrian was looking across the road and wasn't on the sidewalk. Basically every warning sign anybody would look for to expect that person to just walk out into the street. So the car stopped, owner said it "slammed on the brakes", and was annoyed it made the safest autonomous choice possible.

I often want to see peoples driving records when they complain about stuff like that. :D
Should it brake hard for a squirrel? Cat? etc... There will be plenty of debates. Maybe just mad max setting for road kill will take care of that. :)
 
  • Like
Reactions: DrtyJrze
We shall see. In any case, if Tesla is sending "city street autopilot" to Early Access, that's a good sign of FSD progress.

It depends on whether that's NoA on city streets, or just allowing AP, as it exists on highways, to be used in the city. Surprise: AP already works in the city, and has since day 1.

The Reddit thing posted above is some guy saying that a friend of his said it would blow people's minds. When I show people EAP, it blows their minds. I want to hear from actual drivers, what features they got and how well they work.

And if they have City NoA that's safe enough, and ready, to release for a week or a day, why only for a week or a day? Why not just release it. And if it's not ready enough to be safe, it would be criminal to release it even for a day.

FWIW, I absolutely love EAP and use it all the time and it still amazes me. I don't think that stoplight recognition would make enough of a difference to get me to upgrade. Since I have to be constantly aware, it's easy to stop for red lights. The step that will get money out of me will be Level 3: When I no longer have to pay attention. Meanwhile, I'm using EAP on certain city streets, depending on conditions. So plain AP on city streets is not even a step up.

I'm waiting with baited breath (whatever that means!) to find out what the "sneak peek" is going to be.
 
It depends on whether that's NoA on city streets, or just allowing AP, as it exists on highways, to be used in the city. Surprise: AP already works in the city, and has since day 1.

It has to be something more than what we have now, since it is requires EA testing and only works on AP3. So I imagine it will be City NOA with traffic light response and turning at intersections.

And if they have City NoA that's safe enough, and ready, to release for a week or a day, why only for a week or a day? Why not just release it. And if it's not ready enough to be safe, it would be criminal to release it even for a day.

Who said it will only be for a week or a day? Presumably, after EA, it will go wide to everybody with FSD and AP3. In fact, it certainly sounds to me like Tesla is preparing to roll out the 2 FSD features on the website (traffic light + automatic city driving) to the AP3 fleet in Q1 2020.

I'm waiting with baited breath (whatever that means!) to find out what the "sneak peek" is going to be.

The guy claims we will get videos of the sneak peak in a couple weeks so we won't have to wait that long to find out.
 
We've all had this experience: You're walking along and you approach someone standing still and facing away from you. Suddenly, without warning, they turn around and start walking without looking. Even people cannot predict people. But once a car learns to tell that an object is a person, it will be able to react more quickly than a human driver, and being fully aware of everything around it for 360°, it will be able to choose the action most likely to avoid hitting the person.

Funny, my understanding is that right now Summon mode can't even avoid stationary objects.
 
Found this post from a Chip Stratton on the Model 3 FB page with some interesting tidbits on what Tesla is working on:

Chip Stratton
Conversation Starter · 5 hrs
I ran into a Tesla Autopilot engineer today, who provided some interesting information:
1. They are continuing to work hard on Smart Summon, and a version he has tested is performing extremely well in parking lots packed with holiday shoppers.
2. Autopilot is not as smooth on the more powerful Model 3 vehicles as on the less powerful ones and they are working on that problem.
3. Early Adopters should get city driving autopilot before year end.
4. Pothole and road debris avoidance will be very difficult using purely neural net techniques so they are exploring alternative approaches to lessen the load on the computer.
5. Reducing phantom braking is currently a very high priority.
6. Tesla gathers huge amounts of data when a vehicle is in autopilot; but little when not.