Separate names with a comma.
Discussion in 'Autopilot & Autonomous/FSD' started by Lex_MIT, Apr 6, 2018.
Done, looking forward to seeing the results.
Done Very good questions, not too long and straight ahead.
You full comment is great. Very insightful. Just to quickly respond to one of the suggestions quoted above. On the TACC front, you're right. The focus of the survey is on Autopilot (TACC + Autosteer) not TACC alone but it's important to clarify that. I added questions about TACC usage and also created/added an image to clarify that we're mostly asking about #3 in the following:
Although the survey questions were very well written, this added level of specificity makes it better. Frankly, this is by far the best survey on this subject I have seen. For that matter it is one of the best on any subject. Proper questionnaire design is not really common ( I used to teach the subject) so I am really impressed when I see it done. This one should be taught as an example of how to discern new technology experiences without resorting to jargon or unduly complex questions. This refinement proves there is no substitute for really knowing the subject if one wants to makes questions simple. Occam's razor works with almost all topics after more than 700 years, doesn't it?
I agree you already had some of my list.
Your “leading vehicle braking, moving slower, or stopped” should be split into 2 or 3 choices IMHO.
1. I virtually never have to take control when a car that has been in front of me slows. AP handles this fine.
2. A car changing lanes to be in front of me is a special case and AP is so moly too slow in its response.
3. A well known problem is AP sometimes doesn’t stop if there is a stopped car in front of it that it did not see slow to a stop. This can happen e.g at a red light where the car in front of you changes lanes to reveal a stopped car. This rarely happens to me as I mainly use AP on the freeway.
I understand if you don’t want to have so many categories, so do what you think is best.
Also, in my example of passing a large truck, my car stayed centered, it did not drift to a side of the lane. But if the truck is too close to my lane, I want my car to move off center in my lane away from the truck.
The other problem is AP/TACC is really only for highway/freeway miles. So asking in percentage of total miles/hours is going to skew the results for those of us that use it properly. I can't exactly use TACC on the way to work because it's all surface streets and stop lights. That's going to cause more work than it's worth. But, it's not a reflection of TACC/AP.
For a lot of us the AP vs. TACC is something we only decide when were presented with whatever situation arises on a freeway/highway. The first question for myself is do I need TACC to protect me from a speeding ticket.
Of course if you specify the percentage of highway/freeway miles/hours then it might not capture the crazy people who use it on surface streets.
So maybe have a few questions of usage pattern for Daily driving, and having a few questions of usage pattern for Trips. I know a lot of us use TACC/AP a lot for trips where it comes in really handy.
The Survey also seems to missing a change in usage pattern over time. How we got to that usage pattern is extremely important for survey on AP
In the beginning I used AP a lot as soon as it was released with version 7 of the SW. I believe this was in Oct of 2015, and I used it around 90% of the time on a trip from Seattle to Pismo beach in California. I experienced very few situations where it would glitch, and that was mostly diving for the exits. Sure there were parts of the trip that I didn't trust AP for like some curves on some crazy highway to Santa Cruz, but it was a good companion for most of the trip.
Over time though my confidence in AP diminished. There were truck lust incidents on the way from Seattle to Portland, and other glitches. So my usage pattern went from 90% on trips down to 10% or so. Where now days it's mostly for completely empty roads or complete grid lock. Where anything else requires more oversight of AP than the energy required without it. I also noticed the decrease in situational awareness when I used AP. So I determined that for myself using AP wasn't worth the added risk.
The question of responsibility might be a good one to ask on a survey like this. To see if people are fully comfortable with 100% of the responsibility when AP is doing most of the driving. When I first got AP my answer would have been completely, but over time my answer changed. Why it changed was because of truck lust. If you're not anticipating it having a car suddenly going towards some massive object is quite startling. Where there isn't a whole lot of correction time, and the correction itself can cause oscillations.
I was thinking of this:
Autopilot Buddy Device Fits Tesla
You did post pictures of people distracted while on autopilot. There have been some accidents because of it. Even if that is the exception rather than the norm, it is a problem. I've never even considered texting while driving, and most people don't, but various states have passed laws about it because some people do it and it has caused accidents. It doesn't take many people doing something dangerous to put everyone else in danger.
When I worked at Boeing, I was in a lab that tested all the electronics (avionics) destined for airliners along with a lab that did human interface design for the flight decks. They had simulators for the cockpits of all Boeing airliners in service. My group was responsible for the hardware and we supported everyone. I worked a bit with the human interface people and took an interest in what they were doing. I've always had an amateur interest in Psychology, so I found that sort of thing fascinating.
Those people were tackling some of the problems that face automotive automation today back in the late 80s and early 90s. Though the automotive problem is much more difficult. With airliners they had a number of advantages going in:
1) Anyone at the controls of an airliner is highly trained and their job is highly regulated. Anyone who is a screw up is probably unemployed. Even the general aviation people are better trained than most car drivers.
2) There are fewer vehicles up there and they are only in danger of hitting one another at the landing and take off phase. For most of every flight aircraft are far enough spaced there is no danger to each other.
3) There are no roads up there and planes can maneuver in three dimensions.
Electronics have advanced almost 30 years since, which does make it possible to crack the problem of ground transportation now, but it's still a difficult problem. I do agree with you that it is a solvable problem, but it's also a very difficult one.
Should those of us who completed the survey earlier take the survey again, since it has changed?
Submitted. Looking forward to seeing the results!
If I may suggest:
1. If a owner have more than one Tesla with different autopilot hardware, might be separate them. Like experience on Tesla #1, experience on Tesla #2, etc. I think this will provide more details to your question without confusion.
2. I have different confidence on different road condition and passengers with me. If I'm driving alone, I relay heavily on Autopilot, but if I'm driving with family, large possibility I might choose manual driving.
Based on the great suggestions, I made some (backward compatible) updates to the survey:
- Included kilometers along with miles.
- Added J.D. to list of example professional degrees.
- Added a question "Since you first started using Autopilot, how has the performance of Autopilot changed with new software updates?"
- Clarified that if you own (or have owned) multiple Tesla vehicles, you should answer for the one you currently use the most.
- Small clarifications added throughout.
If you took the survey, you don't need to take it again. The changes and additions are small and don't invalidate your survey response.
Please do let me know if other ideas for improvement come to mind, before I scale it up to more Tesla owners.
Nope, you're good (unless you really want to). For now all additions have been minor, and don't invalidate previous survey responses.
Wow, thank you for the kind words. I am often surprised how quick people are to formulate survey questions about technologies they don't fully understand or have personally experienced extensively. Each of the questions asked here have a lot of data behind them, and will help uncover further insights about how to build safe and enjoyable semi-autonomous systems. I look forward to hearing from a lot of Tesla owners and soon releasing the results.
One of the things I think is a weakness with Tesla's display is the warning messages all pop up at the bottom of the screen. For me, the most comfortable setting for the steering wheel height covers the lower 1 cm or so of the instrument screen and I usually miss the autopilot messages when they pop up. I only notice when the screen starts in with the whole border flashing. I also miss a lot of other messages too.
If the warnings were at the top of the screen, they would be much more visible.
What do you think would be a good question to ask about the instrument cluster?
Maybe you could ask what reminder (message, flashing border or auditory) catches the drivers attention usually. Or to put in percentages on which reminder driver reacts accordingly.
Done - looking forward to the results!
Not sure I would see them ... I don't see that my lights are on beam and other stuff at the top of the screen (because of where I have the steering wheel). Configurable Top/Bottom would be fine of course, and do-able ...
Perhaps solved by M3 having the screen-at-the-side?