survey - FSD autopark experience

[m25nyc]

Hi. I've had a mixed experience with FSD autopark and spent some time on the forums looking for hacks to optimize its ability to detect a space and successfully complete parking, emphasis on the former (if it detects, usually it completes OK).

I thought I'd start a new thread AND create a Google Form that can allow a much more structured, in-depth survey, and, share the Google Spreadsheet that goes with the form so everyone here can benefit from what people share. Maybe someone with more statistical chops than me can analyze it if we get a lot responses and post the answers on this thread.

Here is the Google form survey:

FSD autopark experience

let's share our experience on how to optimize autopark; if you are busy you could just do the first question which is entirely open-ended, remember the SUBMIT button at the bottom of page! here is a read-only link for the spreadsheet please feel free to perform a statistical analysis and share...

docs.google.com

Here is the spreadsheet with results:

FSD autopark experience (Responses)

Form Responses 1 complete this survey via this link:,<a href="https://docs.google.com/forms/d/e/1FAIpQLSchM09E_qfHpnD_6RX-iqkce2EifEsxBdp5Vg_bo0gWKI8KuQ/viewform?usp=sf_link">https://docs.google.com/forms/d/e/1FAIpQLSchM09E_qfHpnD_6RX-iqkce2EifEsxBdp5Vg_bo0gWKI8KuQ/viewform?usp=sf_link</a> share...

docs.google.com

 

[m25nyc]

I will be deleting the sheets and survey form about a week after starting - deleting 10/11/2022 - if no more than 10 responses on optimizing parking. So far, only one - me! As FSD beta regularly updates, we would need to attract at least a few responses every month or so in order to help drivers optimize autopark.

 

[Scott Norberg]

I have been fighting with AutoPark, and I just assumed I was missing something in the documentation. So a week ago I took my Tesla x 2022, 2022.28.2 (with enhanced autopilot) to a empty parking lot with reasonable lines. I was able to get it to park in some parking slots but these were were angled slots. I think I had read that they need to be perpendicular for AutoPark to work. But I had some success, but nothing that made me feel confident about using it for real.

Yesterday I was at a supercharger and I thought that would be the perfect place to try this out. Wrong, my X would not indicate any of the slots as valid for parking, at a Tesla SuperCharger, really.

So either I don't understand or this feature is severely broken.

It would be interesting to hear other peoples experiences with autopark before I give up on it.

 

[momo3605]

I will be deleting the sheets and survey form about a week after starting - deleting 10/11/2022 - if no more than 10 responses on optimizing parking. So far, only one - me! As FSD beta regularly updates, we would need to attract at least a few responses every month or so in order to help drivers optimize autopark.

I looked the sheet. Way too many free form responses and questions. Too much work haha. I suggest keeping it simple.

 

[DCGOO]

I have been fighting with AutoPark, and I just assumed I was missing something in the documentation. So a week ago I took my Tesla x 2022, 2022.28.2 (with enhanced autopilot) to a empty parking lot with reasonable lines. I was able to get it to park in some parking slots but these were were angled slots. I think I had read that they need to be perpendicular for AutoPark to work. But I had some success, but nothing that made me feel confident about using it for real.

Yesterday I was at a supercharger and I thought that would be the perfect place to try this out. Wrong, my X would not indicate any of the slots as valid for parking, at a Tesla SuperCharger, really.

So either I don't understand or this feature is severely broken.

It would be interesting to hear other peoples experiences with autopark before I give up on it.

Perpendicular or Parallel is required. It does not support angle parking. It used to require a single spot with two cars immediately adjacent. Now though, Tesla Vision can use the cameras to identify available slots with painted lines on the pavement, with no other cars nearby. It works fairly well, but it is terribly slow. The problem is knowing it has found a spot. It is at the top of shift slide area on the main screen, and is pretty subtle. I think it changes from the "P" to "Park now" or somthing to that effect. The few times I've tried a parallel park maneuver, I found it to be quite good. But I rarely get the opportunity to use it. Everything around me is Perpendicular (parking lots).

 

[m25nyc]

I looked the sheet. Way too many free form responses and questions. Too much work haha. I suggest keeping it simple.

I agree. I started simple but realized there are 2*2*2*2*2 = 32 major combinations for which different distances away from space and in front of space may apply and anecdotal responses unless identifying which of the 32 combinations it applied to, would't be helpful.

head in vs. parallel
on the right side only (a single line of head-in / single side of the street for parallel) vs. on both
one-way traffic flow vs. both-directions traffic flows (in parking lot or on street!)
one-side only spots are driver side vs passenger side
both-side spots car is driving closer to driver side vs closer to passenger side (one-way traffic you could pick)

Anyway, I've deleted the spreadsheet as no one responded.

My guess is that Tesla engineers are able to capture all the above and many more variables from data they already gather. The fact that autopark is so unreliable in detecting spaces, suggests either a very low priority to improve, or, that it is a very difficult problem to address with current vehicle sensors, or both.

 

[m25nyc]

Hi. I've had a mixed experience with FSD autopark and spent some time on the forums looking for hacks to optimize its ability to detect a space and successfully complete parking, emphasis on the former (if it detects, usually it completes OK).

I thought I'd start a new thread AND create a Google Form that can allow a much more structured, in-depth survey, and, share the Google Spreadsheet that goes with the form so everyone here can benefit from what people share. Maybe someone with more statistical chops than me can analyze it if we get a lot responses and post the answers on this thread.

Here is the Google form survey:

FSD autopark experience

let's share our experience on how to optimize autopark; if you are busy you could just do the first question which is entirely open-ended, remember the SUBMIT button at the bottom of page! here is a read-only link for the spreadsheet please feel free to perform a statistical analysis and share...

docs.google.com

Here is the spreadsheet with results:

FSD autopark experience (Responses)

Form Responses 1 complete this survey via this link:,<a href="https://docs.google.com/forms/d/e/1FAIpQLSchM09E_qfHpnD_6RX-iqkce2EifEsxBdp5Vg_bo0gWKI8KuQ/viewform?usp=sf_link">https://docs.google.com/forms/d/e/1FAIpQLSchM09E_qfHpnD_6RX-iqkce2EifEsxBdp5Vg_bo0gWKI8KuQ/viewform?usp=sf_link</a> share...

docs.google.com

I've deleted the spreadsheet as no one responded - replies suggest it was too complicated.

 

[m25nyc]

Perpendicular or Parallel is required. It does not support angle parking. It used to require a single spot with two cars immediately adjacent. Now though, Tesla Vision can use the cameras to identify available slots with painted lines on the pavement, with no other cars nearby. It works fairly well, but it is terribly slow. The problem is knowing it has found a spot. It is at the top of shift slide area on the main screen, and is pretty subtle. I think it changes from the "P" to "Park now" or somthing to that effect. The few times I've tried a parallel park maneuver, I found it to be quite good. But I rarely get the opportunity to use it. Everything around me is Perpendicular (parking lots).

It's interesting I use FSD beta autopilot on highways with very weak lane markings, and vision detects the lanes superbly. So well, that it keeps dead center even when other drivers in cars next to me that I know don't have steering assist options start drifting to the edge of their lane (I wish Tesla would move further away!). But in parking lots, unless 2 cars either side or very very clear lines, no luck with autopark detection. I also find it is quite slow to engage and, that you have to be beyond the space and back in, confuses drivers behind me in crowded parking lots who are impatient and at places like Costco, you want to be head-in not back-in as you need to load the car when you leave and it so crowded, there is no way you could pull the car out and then load it.

Since the car drives itself forward at 80MPH very well, I wonder why it can't park head in? Probably because of the angle of front cameras / optimized for highways? Tesla seems to not be soon adding a whole bunch more cameras or other sensors and so I don't expect autopark to improve much.

 

[vapor trail]

Just my observation, but seems to me self parking has markedly degraded with updates. Not sure when change occured, but when I first had my 2019 Model S, autopark was fairly reliable at seeing spots and did really good job parking centered in the space. Now, it's rare for my car to even see available parking spots, to the point where I've stopped trying to even use it.

 

[kabin]

Just my observation, but seems to me self parking has markedly degraded with updates. Not sure when change occured, but when I first had my 2019 Model S, autopark was fairly reliable at seeing spots and did really good job parking centered in the space. Now, it's rare for my car to even see available parking spots, to the point where I've stopped trying to even use it.

You are correct. There are youtube comparison tests showing that as well as how much better other car manufacturers autopark. As you said, early Tesla model years worked better than current model years. The current model years refuse to acknowledge parking spaces and failed to parallel park anywhere close to the curb. As crazy as it sounds sometimes half of the parked car remained in the roadway.

Tesla may still be struggling as Musk said the team's deadline for autopark and smart summon was the end of Sept.

 

[m25nyc]

You are correct. There are youtube comparison tests showing that as well as how much better other car manufacturers autopark. As you said, early Tesla model years worked better than current model years. The current model years refuse to acknowledge parking spaces and failed to parallel park anywhere close to the curb. As crazy as it sounds sometimes half of the parked car remained in the roadway.

Tesla may still be struggling as Musk said the team's deadline for autopark and smart summon was the end of Sept.

It is strange because FSDb driving on city streets and highways / freeways is so much better than any other car manufacturer and keeps getting better. I was just reading about how Mercedes is boasting their S-Class will fully self-drive "on highways up to 37 MPH" (in Germany only). I was once driving a VW with traffic aware cruise control that turned on/off above/below appx 40MPH and it was incredibly dangerous!

 

[m25nyc]

You are correct. There are youtube comparison tests showing that as well as how much better other car manufacturers autopark. As you said, early Tesla model years worked better than current model years. The current model years refuse to acknowledge parking spaces and failed to parallel park anywhere close to the curb. As crazy as it sounds sometimes half of the parked car remained in the roadway.

Tesla may still be struggling as Musk said the team's deadline for autopark and smart summon was the end of Sept.

I've also posted on the thread about ultrasonic sensors being removed, the USS are extremely helpful for manual parking.

 

[vapor trail]

I've also posted on the thread about ultrasonic sensors being removed, the USS are extremely helpful for manual parking.

Glad my car has them. will be horrified if Tesla brain trust disables them (like they've done with radar) in future "updates"