Enhanced Summon makes FSD seem light years away

[Brettski]

I'm normally "all in" when it comes to Tesla, new features and the promise of FSD.

However - I've tried enhanced summon a few times now - in areas with no traffic and no unusual obstacles. I'm sorry to say it has been an epic failure each time - not once did the car make it close to the target....each time I had to get in and drive it to safety. Had I not aborted, the car would have hit the garage wall at home, and adjacent cars in a parking lot.

So my point here is - if the thing can't drive 100 feet on it's own - how is it ever going to safely drive on real roads?

I'm looking forward to being proven wrong.

 

[zanary]

What build of v10 do you have? There are been four builds to date that we know of

If you look at all of the YouTube videos, they look pretty successful, only a few minor accidents or close calls.

 

[gigawatt1010]

Most of the Enhanced Summon videos I’ve seen are Model 3’s. Is it possible that the M3’s are better suited or have better equipment for Enhanced Summon?

My MX is installing v10 as I type. Can’t wait.

 

[Brettski]

What build of v10 do you have? There are been four builds to date that we know of.

I have 2019.32.11
I haven’t watched the YouTubes.... but I tell you, I won’t attempt it again if there’s even one moving car in the vicinity...... it was that bad. It got perilously close to everything around it. (Such as the car parked next to it, the curb..... and in general it just never managed to align with the usual direction of traffic flow.)

 

[croman]

My experiments have gone similarly. Row of parked cars, disabled sign post, and my garage. It went for all 3 when there was no reason. Also it messed up very suddenly. Not worth it.

 

[MarcusMaximus]

Hmm, I’ve used it near my house a bunch(mostly when I needed the car out of the way to move stuff in the garage) and once so far in a parking lot, all pretty successfully. In the parking lot, it even went wide around a lady loading stuff in her car from a shopping cart.

 

[cucubits]

Can't help but find it a bit funny that only now some start realizing this about FSD.

Elon sure is a great car salesman, had good marketing ideas or some ppl are really gullible and willing to pay for nothing.

 

[jebinc]

I tried it only once, “once....” - movie, “Johnny Dangerously”....

 

[BPeter]

Can't help but find it a bit funny that only now some start realizing this about FSD.

Elon sure is a great car salesman, had good marketing ideas or some ppl are really gullible and willing to pay for nothing.

Only a matter of time now before the facade comes crumbling down. Back in the beginning of this year when I was deciding what to buy I just did a few hours of research of AP1 vs AP2 and AP2.5 and concluded that none of those systems will ever be FSD, so I decided to just buy a used AP1 car. But then the firesale happened and I bought mine with AP2.5. I still believe I would've been very happy with AP1, other than not having dashcam.

 

[dannycamps]

I tried it only once, “once....” - movie, “Johnny Dangerously”....

It's fargin warfare!

 

[cwerdna]

Can't help but find it a bit funny that only now some start realizing this about FSD.

Elon sure is a great car salesman, had good marketing ideas or some ppl are really gullible and willing to pay for nothing.

Indeed. I and others have harped on the optimism of some folks (and fanboys) here and a few other forums re: Tesla's timelines and "promises".

We can point to Tesla's missed dates. I keep pointing them to the efforts that Waymo has had for many many years, their robotaxi efforts in Phoenix, CA DMV autonomous testing disengagement reports, what Cruise Automation has showed and what even Nissan (not a leader in this space) has shown years ago to a reporter yet some folks persist...

If Tesla demonstrated parity or exceeded the above players, then the optimism would have more validity.

(For those who don't know what some of what I'm referring to, see these:
UPDATE: Disengagement Reports 2018 – Final Results
Tesla's vs. Google's/Waymo's vs. Cruise Automation's results at Testing of Autonomous Vehicles.
Why testing self-driving cars in SF is challenging but necessary

- from 2017
Cruise - take a look at some of their San Francisco videos, the 1,400 left turns video and the double-parked one

from 2012)
And, this is not to say Waymo's close to perfect judging by reports we hear out of Phoenix.

 

[emmz0r]

I experienced a quite severe regression in AutoSteer. Finally in 2019.32.2 they had fixed the problem that the car wants to throw me in different bus stop pockets. Now it tries to throw me into the crash barrier on the side before the bus stop...

 

[u00mem9]

Hmm, I’ve used it near my house a bunch(mostly when I needed the car out of the way to move stuff in the garage) and once so far in a parking lot, all pretty successfully. In the parking lot, it even went wide around a lady loading stuff in her car from a shopping cart.

that lady is an innocent bystander, and certainly not a willing participant in this experiment. Maybe if we explain how crucial the share price is to Tesla’s long term viability, she’ll understand... but I don’t think it’s acceptable to test the system when others are around. Easy to imagine things going wrong even with the dead man switch in hand. Please reconsider.

 

[whitex]

Many people, myself included, have been warning/preparing FSD customers for a while. Elon told us "AP1 Summon will find you anywhere on private property", actually delivered "car will drive up to 40 ft forward or backward in a straight line while the drive holds a dead-man-switch to ensure the car doesn't hit anything". For AP2.0 Elon promised ability to summon the car across the country. So far, AP2.0 summon is the same as AP1 summon. For AP2.5/AP3, slight improvement, the car can make it across an empty parking lot while the driver holds a dead-man-switch. Hower, 3 years later, this is not even in the same galaxy as summoning from New York to L.A. I think Elon is closer to reaching Mars than he is to summon across the country.

Btw, enabling smart summon in its current form may backfire, if there is a big enough incident, it may get banned and Tesla might have to legally disable it. Or, maybe that is Elon's plan, so he can say the reason why you can't have FSD is because it's illegal.

 

[jeffdom1978]

Mine has worked great so far! The only time it failed was when someone was standing right behind my car and I didnt know it. It said "Summon Not available" I havent had a chance to bring anyone with to film the attempts. One thing I noticed is if the parking slots are not slanted enough and wide it will go down the wrong way sometimes, but hey so do I.

 

[ramonneke]

Keep in mind that navigating around parking lots likely is *WAY* more difficult than any regular road due to so many dynamics.

Any person can drive a car around the city but ask them to park and then see how many fail to correctly park the car.

 

[emmz0r]

Why are you using enhanced summon around bus stops?

? I Said AutoSteer. I don't even have Smart Summon

Anyway, seems to not distinguish between grass and tarmac:

And in true reddit spirit, it's with a positive spin: Roads... where we're going, we don't need roads! : teslamotors

 

[BigNick]

I think Elon is closer to reaching Mars than he is to summon across the country.

In space, there are fewer obstacles to crash into.

edit: snip quote

 

[MarcusMaximus]

that lady is an innocent bystander, and certainly not a willing participant in this experiment. Maybe if we explain how crucial the share price is to Tesla’s long term viability, she’ll understand... but I don’t think it’s acceptable to test the system when others are around. Easy to imagine things going wrong even with the dead man switch in hand. Please reconsider.

Ditto, perhaps for even driving the car in any form. After all, look at all those Tesla fires! Could explode and kill someone else!

 

[davecolene0606]

OP and some follow on posts here seem to be carrying this to an extreme. So far I have thrown some pretty complex problems at summon with mixed results. It needs supervision and I can see where HW3.0 will do nothing but help. One has to do a bit of pre planing for a successful outcome of a summon. One thing it seems good at is obstacle avoidance. In my trials it seems to lack the ability to track direction flow and this is where planning helps a bit. Having other vehicles around actually improves its decision making as it appears to take their directional flow into account and it conforms to that. In short this update shows a clears path to the goal

AP functions have significant improvement in smoothness and execution, oncoming traffic, line recognition, vehicle recognition, vehicle acceleration and stopping have improve as well.

V10 shows a clear path forward to FSD. HW3.0 is definitely a requirement. I have a bit of concern as to whether MCU1 will be up to the task. Having said that, it appears the MCU interaction is limited to navigation and as long as the MCU isn't bogged down with activities like web pages that require continuous data downloading, it seems to hold together.

V10 2019.32.11, HW2.5 MCU1

Fire Away!