-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
chipmunk
Member
It doesn’t seem to follow this bias at other parts of the video. Look at 4:31. The car stops right in the middle...much further to the right than the car in front of it.
The outboard bias in low speeds is a stated feature and fact from Audi. And makes sense to anyone who drives in Germany. That said, what follows is speculation since obviously user experiences with the system as still limited:
I would guess it not hugging the outboard side earlier because earlier being in the middle of queues (where lead cars are in the middle) it sees no reason to.
It is the lack of queue on one side (that initiates the ending) and slowing speed together that makes it assume the bias. Also were it need to stop for lack of user reaction after the requisite number of seconds, the outboard lane edge is understandably preferred as well...
Again, read my comment about yellow truck in right lane. If the truck distance remains same To camera ( it sure seems close as before) it's unlikely that the lane on right had a gap, hence me calling B's.
Why would the yellow truck be relevant - it is in front?
What is relevant is what is happening to the side and back of the car. If the adjacent lane queue disappears there (e.g. car behind the yellow truck slows down and a gap appears) that would end the scenario. That is what the guy on the video claims happens.
Now, we don't see that of course due to no video existing of it, but we also don't see it not happening. Lacking evidence, I think the simplest explanation is there was a gap...
chipmunk
Member
I’d be more impressed if they said it was a bug they would work out before delivery. In practice I’d be annoyed if that was expected behavior. I would want it to fall back to level 2 control rather than asking me to take over.
This is a perfectly good argument - the limits are conservative and the usefulness of the feature can certainly be debated. Audi will probably eventually lift some of the limits, but who knows when. E.g. Tesla still has time to catch up...
The upside is, when Audi gets legal approval, you can read a book etc. in slow queued up traffic...
Also Level 2 fallback is interesting and too little is known about how that might work. The car obviously has plenty of Level 2 aids as well, but how all this meshes and how Audi tackles mode confusion is interesting...
I am a Tesla and Audi owner. Long history with high-end Audis, previous daily driver before Teslas an Audi A8. I have a Tesla with AP2.
Objectively speaking, current auto-steering systems are like this:
1. Tesla AP1
2. Volvo ProPilot 2.0
3. Tesla AP2
...
6. Audi Level 2
But the thing is, within its scenario, I expect Audi Level 3 to be number 1 on that list. For Level 2, GM's Supercruise may be number 1 when it comes.
Tesla's AP2 is objectively such a mixed bag (and downright dangerous at times) that who knows if and when it will go up or down..
There is some objectivity for you from an Audi guy.
I do think Audi has been impressive, and continues not to rest on their laurels, and if Audi supplants Tesla atop the autonomy heavyweights it will be because of their discipline and persistence, insert German jokes here...
It's easy to feel like Tesla is stagnating because everyone is chasing them now, and they are a slightly easier target because of the AP2 to AP1 parity issues and Elon being a little too overpromising....
What do you of the following scenario?
1) mid-2018 AP1, AP2 and AP2.5 will be virtually at parity and at the front of the pack, basically meaning that we will not be extracting extra performance out of the hardware yet?
2) BMW, Audi and Mercedes start to separate from each other but all from a kind of pragmatic and careful sort of approach... in other words, I see them as being more cautious and erring more on the side of not having you tube videos of owners riding in the back seats
3) some dark horse, in China perhaps, is quietly working on things now in a quiet and disciplined fashion? I'm thinking of the Go computer finally winning....
I find it hard to imagine Tesla not staying in the lead for a long time.
Everyone forgot about current leader in autonomous - waymo. They've been doing lvl 4 demos for a while so if anyone will supplant Tesla and soon, it will be them.
Put it this way: Why I think of Audi what I think, is because I sincerely believe they have had a successful parallel program in autonomous development, one that is independent of what they have been shipping in driver's aids over the past ten years. The new Audi A8 is the first car with this new system.
From this stems two of my opinions: 1) Whatever anyone has experienced in an Audi so far means very little, it is just a completely different system they are shipping in the new Audi A8 and beyond. 2) I have a hard time seeing Audi "chasing" anyone, since they were doing this when Tesla was still fiddling with the Roadster. Audi is one of the autonomous pioneers, even within the Volkswagen group.
There are areas where Tesla has out-innovated and out-pushed others - BEV drivetrains is one. But I just don't think Audi vs. Tesla on autonomous is one of these. Audi has been following their own strategy, very methodologically, and all signs IMO are, started before Tesla did, because they've been in public for so long and private work must have preceded that, possibly going beyond the formation of Tesla...
1) Possible.
2) I haven't followed BMW's latest on this too much, but I believe Mercedes is following in Audis footsteps quite soon - they too have been working methodologically for this for a long time. 2018 for Audi will be about the 37 mph Level 3 system in the Audi A8. Maybe in 2019 we'll see the upgrade to motorway speeds with limited scenario Level 4 (motorways and parking lots) for Audi. These cars probably will still not have FSD capable suites, in the way the Germans understand them, so limited to just some scenarios (Audi's plans outlined here: First L3 Self Driving Car - Audi A8 world premieres in Barcelona). This means there certainly remains a big window of opportunity for Tesla with their 360 suite.
3) Dark horses are always a possibility, perhaps in collaboration with some other company. Interesting to see what Apple does...
That is an OK opinion IMO, based on the aggressive hardware suite and method Tesla has been using to get products to market. Others will probably not ship 360 vision processing (let alone their 360 lidar + radar) for the year or three... That said, the picture may change if we talk about driver's aids vs. self-driving. Do you see Tesla saying "you can read a book, sleep in your car" very soon? I have a harder time seeing that, even if Tesla were to release some sort of FSD beta...
There is also the big if: What if AP2/AP2.5 turns out to be inadequate for FSD as understood by Level 4-5...?
Indeed, aka Google. That is another company that has had a very disciplined approach to this, albeit very different from the Germans. If the likes of Audi have had any inspiration, it is more Waymo/Google, than Tesla IMO on autonomous.
lunitiks
Cool James & Black Teacher
JeffK
Well-Known Member
I wish we could see some behind the scenes stuff at Tesla like this.
lunitiks
Cool James & Black Teacher
I like the idea of not using humans to do labeling. Instead, build photo-realistic road signs, vehicles, debris etc. into the simulator, equip your fake "car" with sensors, drive around and feed the sensor input to your real-world ECU. (The ECU wouldn't care if the input comes from real life or some computer space.) No need for a person to tell if it was right or wrong in classifying the object - the simulator knows exactly what it should have seen, anywhere at all times.
(What's really neat, is that you could test sensor types, brands, specs, placement etc., in the sim.)
Makes sense that Tesla's hiring ppl for simulation purposes:
Question is; is Tesla ahead or behind of the game?
(What's really neat, is that you could test sensor types, brands, specs, placement etc., in the sim.)
Makes sense that Tesla's hiring ppl for simulation purposes:
Question is; is Tesla ahead or behind of the game?
I think you just found the best information on how Tesla is thinking with respect to it's autopilot to date! When in doubt, look at the job positions they are trying to fill.... love it.
lunitiks
Cool James & Black Teacher
I´d dare to say that Tesla actually never was "in the lead" from a technical standpoint.
To be blunt none of the classic manufacturers would ever dare to make their drivers beta software testers.....Ap1 and 2 are Lvl 2 for a very good reason..because they **** up from time to time, and in the earlier versions barely ever reminded the driver about their beta status.....
Imagine the hard and software that the Model S that didn`t recognize the crossing truck had in several hundred thousands or even millions of Audis/BMWs/Mercedes etc...
Seriously..you can`t get away with stuff like that when you`re a big player.
Tesla did it because they were small and needed something special to get the attention of investors and customers.
Last edited:
Similar threads
