The car "sees" further than what is visualized on your IC. Often if cars are ahead driving slow and I'm coming up on them, even before I see any cars depicted on IC the car will begin to slow. That has been my experience thus far.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Elon: "Feature complete for full self driving this year"
- Thread starter dckiwi
- Start date
The car "sees" further than what is visualized on your IC. Often if cars are ahead driving slow and I'm coming up on them, even before I see any cars depicted on IC the car will begin to slow. That has been my experience thus far.
I get the use of maps in the short term. Depending on the level of development, it is replacing a bigger problem (greater inaccuracy) with a smaller problem (lower inaccuracy). However, if the goal is to eventually minimize inaccuracy, then even that better option may not be best.
The basic issue I see with map as ground truth:
Are you, and the car, assuming the map is always correct?
If the map is fallible, how does the vehicle determine that?
In the case of city AP, yes Tesla is using the map as a gating function on stop light detection. @Bladerskb , I was wrong in how it was currently operating. However, they are not gating the detection algorithm, just the resultant action. Assuming they can get the detection level up, I would expect the gating to be removed. However, I would also expect that they may retain the map as a fail safe, resulting in the union of the false positives.
Example why this is good: driving in an area with a blackout. No traffic indicators anywhere. Very helpful to know where intersections are. This is like the difference between a tourist vs a native to the area driver.
Interesting. What about the whole idea that as the accuracy of perception increases, in effect we can start creating our own HD maps in real time as the car drives around. I know Green mentioned that Uber was doing that if I remember correctly. I guess once perception gets to that level, and now is more accurate then the database of HD maps, do we essentially not require HD maps anymore. Ultimately we will go off what is most accurate. And that is why we have all the debate, as there is much controversy surrounding whether our current hardware setup will take us to this level of perception.
Great, then when your car has no signal because the mast is down it can't self drive any more. Your summon from 1000 miles away gets stuck and you are screwed.
Turns out this self driving thing that they have sold you is probably a bit more difficult than people generally assume.
MrBadger
Badger out
Most traffic lights have local controllers, so there is no central server as such that knows what the traffic lights are up to exactly although the traffic control centre may get some batch feedback as to what it has been up to, but its very granular compared to what is needed by a vehicle to know upcoming sequencing thresholds. So there is no centralised resource to get this information.
My understanding is that the information from V2I is via localised beacons (I believe 5G was the most likely candidate) and that global standardisations and security is/was (several years since I got any insider knowledge) a slow burner. I suspect part of it was initially down to coming up with a solution to what was not yet a widespread problem that needed solving. I do know that even 3 or so years ago, our traffic light controllers would have been pretty much 'feature complete', but there is a long way between that and a commercial partner taking that feature and building it into a real validated product that met global standards and the vehicle manufacturers then making use of that feature in their vehicles.
I worked on an air quality project where we took data from traffic lights (often several minutes old due to above limitations) and then used this information to simulate a number of future traffic flows and emissions and come up with a set of air quality models for the forth coming time frame. We could then adjust traffic light sequences to balance traffic throughput and air quality. Unfortunately the project never made the light of day due to a number of issues that would have been nigh on impossible to resolve into a multi stakeholder commercial product. So I know how hard it is to work on systems connected to real traffic control systems.
It's not like the REST API is required to drive. It's more a way to optimize traffic when it's working.
Sounds like the traffic control industry has the same challenges as the rest of the automation industry. Outdated equipment and in general little IT knowledge.
I'm a tech-lead in a company that develops cloud services for automation industry. We see that there's a lot of potential going at this industry the Tesla way. Every controller, like the traffic light controller, should get software updates from the cloud automatically
. We've done this in pretty much every other part of the automation industry, I guess we should try to look at traffic lights too
MrBadger
Badger out
Don't underestimate their knowledge. Its a very dangerous thing to do to underestimate how hugely complex things work, especially if thinking a few lines of code may be all that is needed. Many instances of people who think they know how to do things only to find out that they don't - normally they don't last long. I was tech lead in a company that developed cloud based solutions too, so I understand the realities. Right tool, right job, right place. There is a huge difference between technical capabilities and what risk is palatable by various organisations.
Oh and yes, we ran our traffic control and air quality system as SaaS. But our second by second safety critical optimisers often situated in the middle of nowhere are best left to their own devices where they can carry on saving hundreds of lives. Right tool, right place. Cloud isn't it.
Oh and yes, we ran our traffic control and air quality system as SaaS. But our second by second safety critical optimisers often situated in the middle of nowhere are best left to their own devices where they can carry on saving hundreds of lives. Right tool, right place. Cloud isn't it.
linux-works
Active Member
It's also how older companies go bankrupt or just become smaller until they disappear. Look at Kodak, Nokia and other older companies, and look at Tesla and SpaceX's position today. A lot of aerospace engineers laughed about SpaceX, now see who's laughing.
We've already caused a lot of disruption in the market by doing things in the cloud, that most people didn't think could be done in the cloud. Cloud is just a tool, with a lot of ways to implement. There's a lot of misconceptions about what you can do and cannot do in the cloud based on experiences with existing cloud solutions.
MrBadger
Badger out
Ah the disrupted word. Many call themselves disruptors but its not what they call themselves, its about what their products and services achieve. The company that I worked for never thought of themselves as disruptors, but when you looks at their achievements over decades, they were and remain disruptors. They have developed products and techniques that thousands of people have to thank their lives for, saved many millions of minutes and will have probably had a positive effect of probably every road user across the planet. They absolutely lead many fields. All, not for profit. Google turned round to us and told us how far ahead we were from them for some of our autonomous vehicle work. So don't go belittling others because you think a few lines of code is the solution to a problem that you have no idea how complex is. I've seen two so called self proclaimed 'disruptors' try to apply their little knowledge to a problem. One, who thought that they could create a traffic control system using neural net were soon replaced (by us) in the consortium and another who said they could build a cloud based traffic control system for a large European city in 6 months have managed to date, 4 years later, a single trial traffic light nowhere near a city. As I said, both called themselves disruptors, but they have done no disrupting whatsoever other than channelling money away from other projects.
Last edited:
diplomat33
Average guy who loves autonomous vehicles
“I am hoping to roll out traffic lights and stops to wide US release in a few weeks and probably a worldwide release in Q3 (so many variations in each country!). Very important to make sure this is done right.” Elon
Tesla is aiming to release Autopilot for intersections in just weeks in the US, will take months for other markets - Electrek
Tesla is aiming to release Autopilot for intersections in just weeks in the US, will take months for other markets - Electrek
linux-works
Active Member
do we know if this is part of the FSD paid package, or current AP or even EAP?
anyone know where this feature falls?
anyone know where this feature falls?
diplomat33
Average guy who loves autonomous vehicles
I heard a rumor that EAP with HW3 will get this feature. But personally, I doubt it. The feature is listed as a FSD feature on the website and it was never an EAP feature. So I would be very surprised if Tesla suddenly made it an EAP feature.
Should be FSD for both EAP and AP and in all likelihood only with HW3.
Edit: @diplomat33 beat me to the punch
linux-works
Active Member
ah, ok. for some reason, I was thinking this was going to be part of regular old AP and EAP, not an FSD-exclusive.
note to elon and tesla: no, this STILL isn't enough to justify many of us paying the $7k for fsd.
I'd beta test it if it was part of AP, but since its not, well, I'll never be experiencing this feature.
you'd think they'd want as much exposure as possible. wonder how many with FSD are willling to opt-in to this?
note to elon and tesla: no, this STILL isn't enough to justify many of us paying the $7k for fsd.
I'd beta test it if it was part of AP, but since its not, well, I'll never be experiencing this feature.
you'd think they'd want as much exposure as possible. wonder how many with FSD are willling to opt-in to this?
willow_hiller
Well-Known Member
You're in for a shock, then. In that same Twitter thread Elon said that the price of FSD will increase in July when the feature has rolled out (likely to $8k).
linux-works
Active Member
elon thinks he can charge more for it, in a down econ?
perhaps he thinks that the well-off are insulated from the economic ups and downs.
"its a bold strategy, only time will tell if it pays off"
he better make his software money now, since a few years from now he won't be the only game in town.
perhaps he thinks that the well-off are insulated from the economic ups and downs.
"its a bold strategy, only time will tell if it pays off"
he better make his software money now, since a few years from now he won't be the only game in town.
willow_hiller
Well-Known Member
I once had a French colleague who was offended that bread cost extra with a meal in Switzerland. The cashier explained "Well, someone needs to pay the baker!"
FSD is definitely out of my price range for the moment, but if Tesla is investing an increasing amount of time and resources into FSD, it makes sense the price will go up too.
I can only hope they offer some sort of pay-over-time plan for those who didn't roll it into their auto loan. A lot of people can stretch into a Tesla with a decent credit score, but not a lot can afford to spend $7-8k outright. I did pay $3,000 for base Autopilot in cash, and even that felt rough despite all the utility it comes with.
FSD is definitely out of my price range for the moment, but if Tesla is investing an increasing amount of time and resources into FSD, it makes sense the price will go up too.
I can only hope they offer some sort of pay-over-time plan for those who didn't roll it into their auto loan. A lot of people can stretch into a Tesla with a decent credit score, but not a lot can afford to spend $7-8k outright. I did pay $3,000 for base Autopilot in cash, and even that felt rough despite all the utility it comes with.
willow_hiller
Well-Known Member
Similar threads
