Tesla Autopilot Software - MIT Tech Review article

[malcolm]

Tesla Tests Self-Driving Functions with Secret Updates to Its Customers’ Cars

“Since introducing this hardware 18 months ago we’ve accrued 780 million miles,” said Anderson. “We can use all of that data on our servers to look for how people are using our cars and how we can improve things.” Every 10 hours Tesla gets another million miles worth of data, he said.

Tesla’s engineers initially test new self-driving software against those records. Any that perform well can also be tested by secretly installing them onto customer vehicles and watching how they respond to conditions on the road, although the software doesn't actually control the car.

Clever development system.

 

[bradhs]

I'm sure they are collecting information on different drive algorithm sets but only have one set as the primary set in use for driving the actual vehicle when in AP mode.

 

[RDoc]

Before they sprain their arms patting themselves on the back they should fix some glaring bugs in the software.

For example, it would be nice if when AP is changing lanes it wouldn't overshoot the lane it was going into. This has been a problem since AP was first introduced and is still very much there. How the system can steer the car so far over that it generates the lane departure warning against it's own driving is completely beyond me. It's particularly hair raising if there's another vehicle in the lane it's overshooting into.

 

[bradhs]

It never overshoots the lanes here in San Diego. Maybe bad road markings where you're at.

 

[Tequesta]

My car never overshoots the lane in a lane change during AP. Any others having this problem too?

 

[bradhs]

Where do you like and what kind of roads is it overshooting? Post a picture of the road, very curious to see what it looks like. Where I live (Southern California) I never have that problem and I use the lane change a lot.

 

[kort677]

I've never experienced an "overshoot" on a lane change, almost 11k miles on AP

 

[CHG-ON]

Mine has never overshot. I give it exactly 5 clicks of the blinker, which gets me 1/2 way into the new lane and let go. It settles in every time. Are you holding it longer? I don't know if that would have anything to do with it.

 

[bradhs]

It doesn't matter if you hold it in longer, it will properly enter the next lane and not go any further. I've never had an issue either, so it's likely something with the lane markings.

 

[animorph]

All these posts on different behaviors just seems to support the report that Tesla tries out different approaches with OTA updates to subsets of the car population.

 

[cronosx]

Before they sprain their arms patting themselves on the back they should fix some glaring bugs in the software.

For example, it would be nice if when AP is changing lanes it wouldn't overshoot the lane it was going into. This has been a problem since AP was first introduced and is still very much there. How the system can steer the car so far over that it generates the lane departure warning against it's own driving is completely beyond me. It's particularly hair raising if there's another vehicle in the lane it's overshooting into.

Maybe they are planning on using the new nvidia tech its a different tecnology so maybe it willfix your problem and probably it will create some new problem that now you dont have .. if they change ... And maybe the news is that they will since it has little sense that you have a super better autopilot and you dont release it. The current problem with ap cant be solved using only new hw

 

[malcolm]

Hi RDoc. I'm unclear if you actually have the overshoot "problem"?

According to your signature your S85D was delivered in April 2015, nearly six months before Autopilot was announced

 

[kort677]

as for the the overshooting, I've just experienced this problem for the first time. I was only able to replicate it once. the car was travelling in the far left, passing an 18 wheeler, after passing I engage the turn signal to have AP pull into the center lane, another 18 wheeler was in the right lane just ahead of where I was headed, the car makes the lane change properly to center lane but then overshoots the lane and is a foot or three into the right lane headed for that 18 wheeler's rear end, like the truck ahead of me was a target to be hit. of course I assumed control. a few miles down the road I replicated that scenario and it did it again. if the right lane is empty it doesn't overshoot. I couldn't replicate the overshoot headed into the left lanes.

 

[Zybd1201]

I have had both the overshoot problem but even more concerning I have used auto-lane change , turned the blinker off when I am 3/4 in the new lane and had the car dart back to the old lane.

I'm really feeling left out here. Do all of these people who have never had problems with AP actually own the same car I do? Are they just being generous to support the cause? I'm all for supporting the cause but the AP has tons of issues.

 

[calisnow]

I have had both the overshoot problem but even more concerning I have used auto-lane change , turned the blinker off when I am 3/4 in the new lane and had the car dart back to the old lane.

I do not have the overshoot problem - 12,300 miles of driving on 7.1 here in So Cal. As for the blinker I leave it on until I'm in the next lane.

 

[royewest]

I haven't had overshoot issues, though some hesitation when it should be clear -- but I can certainly live with that.

Anyone have to manually turn off the blinker after the lane change? I have to every time. It's on my dozen-or-so-long list of little things for my first SC visit.

 

[thimel]

I haven't had overshoot issues, though some hesitation when it should be clear -- but I can certainly live with that.

Anyone have to manually turn off the blinker after the lane change? I have to every time. It's on my dozen-or-so-long list of little things for my first SC visit.

Yes, one always has to manually turn off the blinker. That is how it works.

 

[Electroman]

Never had the overshoot problem.

 

[royewest]

Yes, one always has to manually turn off the blinker. That is how it works.

Oh, thanks for that. I'll cross it off my list!