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

Has anyone used trash cans etc to test the limits of Summon?

This site may earn commission on affiliate links.
So as many people are exploring with the new advanced summon, I've had a lot of close calls on my very limited tests, and I don't know whether the close calls were not "close calls" but merely me not trusting in the technology to do what it should do.

I, being just to the side of an aisle of cars my car was pulled into, used the advanced summon. It started backing out flawlessly.. until it had turned far beyond the point it needed to as it backed out, in order to stop, turn the other direction, and pull forward towards me. When it got within a foot of the parked car on the other side and didn't look like it was going to stop, I let go of the button and it stopped.

I re-positioned myself so I had a better view of just how much clearance it had, and then attempted the summons again. Expecting it to pull forward, instead, it continued on it's rearwards traverse, with me immediately stopping it. It stopped 6" from the bumper of the car behind it. At that point, I decided it was too much of a risk and walked to it.

Being computerized, and understanding feet and inches, had I not chickened out, would it have stopped itself an inch from the other vehicles bumper, than turned out, or was I right to stop it where it was? I haven't tried setting up trash cans to simulate the process and see if it hits the trash cans, but a car seems like a large enough, blatantly obvious object to recognize, especially in broad daylight.

I tried it again later in the day, where I had backed into a parking spot, to see if it worked better. It did. It pulled out and started driving towards me, until it saw the pedestrian walking towards the store I was at, close to the side of vehicles. Ist stopped about 10-15' short of her and stayed in place. I kept the button pressed, and eventually, after she walked another 3-5 feet, it pulled forward about 1 more foot, then stopped. Then it pulled forward again about 6 inches, and stopped a third time, with an error popping up on my phone stating "error, pedestrian" and failed.

It's amazing it can see a pedestrian 15' away and think it's in danger of hitting it, but it can't see a car 6" from it.

My last attempt was at night in a parking lot with nobody around. There was a planter between myself and the vehicle, and the curb was painted red. It pulled out of the parking spot and started coming towards me, then suddenly it turned hard right and started turning straight for the curb. I stopped it about 6"-1' from the red painted curb with the car tires turned hard right. Had I not stopped it, it would have most definitely hit the curb!

Have any of you used trash cans or cardboard boxes/other items that wouldn't hurt the car, to test exactly how well it detects objects? Would love to hear your results.

So far that I can tell, it's an extremely scary feature to use, risking in my case, a $60k car to save from walking 20-30 feet. I can't imagine Tesla repairing my car for free because I didn't let go of the button in time and it slammed into a wall that it should have seen! I'm scared to use the feature at the moment!
 
I’ve had both successful tests and failed/too scary to let it continue tests.

From a parking space, it has successfully pulled out, driven past maybe 10 other parking spaces (with cars in them), paused at the “intersection”, turned right, then immediately turned right down a parallel aisle (up a hill) to where I was standing and watching the whole time.

It has also pulled out of a parking space (all parking spaces filled, but no real car/pedestrian activity) and ”come to me” to pick us up from a wedding reception (wasn’t that far and could see the vehicle the whole time).

But I have also seen how it almost hit a wall in a parking garage (doesn’t seem to know multilevel parking garages yet, just routing in a 2D plane I think). Another failed test the app showed the direction opposite of actual vehicle orientation (meaning it was facing pretty much south but in the app it looked like it was facing north). Starting smart summon it started heading towards a wall and not me and the app seemed to show it moving the opposite direction. So something was confused about the orientation of the car and moving forward in drive was actually showing a less and less accurate positioning of the car since it was moving in the opposite direction that what it thought it was doing. Unsure whether this was an issue in the app reporting inaccurately or the car was actually wrong.
 
Yes, it is. If human brains don't need LIDAR, then computer brains don't either. The brain just has to be better than it currently is.

You do realize that modern day humans have been on this earth for several hundred thousand years, right? Across roughly 7 billion people, that is an obscenely long time to learn how to interact with the world around us. Do you expect a computer to just magically start to understand the world in the next 3 months maybe, 6 months definitely just because Tesla has a couple computers whirring away in some data center somewhere?

This is the most absurd argument that I see bandied about on a regular basis.
 
  • Like
Reactions: SandiaGrunt
You do realize that modern day humans have been on this earth for several hundred thousand years, right? Across roughly 7 billion people, that is an obscenely long time to learn how to interact with the world around us. Do you expect a computer to just magically start to understand the world in the next 3 months maybe, 6 months definitely just because Tesla has a couple computers whirring away in some data center somewhere?

This is the most absurd argument that I see bandied about on a regular basis.
Humans have been playing chess for hundreds of years and yet computers are already orders of magnitudes better than humans, after having played for just 4 hours, against itself.

Let the computers train its brain over billions of miles in a few years, more than any one human could in a lifetime, and I'm sure it'll be just fine.
 
Humans have been playing chess for hundreds of years and yet computers are already orders of magnitudes better than humans, after having played for just 4 hours, against itself.

Let the computers train its brain over billions of miles in a few years, more than any one human could in a lifetime, and I'm sure it'll be just fine.

Yeah, because playing chess is totally analogous to maintaining a full working understanding of the world while flying down the road in over 2 tons of metal at speed. :rolleyes:

The funny thing is that I actually agree that at some point in the future, vision only autonomous driving will actually be viable. The catch is that this will not be in the lifetime of any Tesla currently on the road.
 
But I have also seen how it almost hit a wall in a parking garage (doesn’t seem to know multilevel parking garages yet, just routing in a 2D plane I think). Another failed test the app showed the direction opposite of actual vehicle orientation (meaning it was facing pretty much south but in the app it looked like it was facing north). Starting smart summon it started heading towards a wall and not me and the app seemed to show it moving the opposite direction. So something was confused about the orientation of the car and moving forward in drive was actually showing a less and less accurate positioning of the car since it was moving in the opposite direction that what it thought it was doing. Unsure whether this was an issue in the app reporting inaccurately or the car was actually wrong.
If there was a concrete or metal roof above your head, there's no way for a car receive an actual GPS signal. The signals are too weak. If the car has an actual compass (magnetometer) aboard, it could be thrown off by electrical equipment and I figure live wires.

Your phone's position estimate may also have been way off or inaccurate due to the above.
 
Lidar's a crutch, right? ......right?
Yes.

Memorizing things by heart may be beneficial in the short run … but won't take you far.

Ok, so why do Tesla cars have radar? Humans don’t shoot cm-wavelength radio waves out of our noses.
Because they are cheap. If Lidar was as miniscule and cheap as radar - I think it would be useful to put it in a mass market car.

The choice currently is between Lidar and big data.
 
Last edited: