Forums

False positive forward collision event

False positive forward collision event

I took delivery my 2016 75D less than a month ago and I love the car. However, this week I had a concerning experience with autopilot (running version 8). While traveling on the highway at night with the nearest car almost out of sight, the dash suddenly displayed a red car and slammed on the brakes. Auto steering was immediately disengaged but cruise control remained active. This event only lasted a second but it was enough to scare the heck out of me. Upon arriving home I emailed my local Tesla service team and they pulled my diagnostics the next day. They said the car detected an obstacle 30 feet away and engaged emergency braking (forward collision sensitivity set to medium). They couldn't explain the false positive event and suggested it could have been "a piece of cardboard or a bird being detected in just the right way to trip the sensors."

Has anyone experienced a similar false positive forward collision event? I do not want to disable a safety feature, but I am afraid this could cause an accident on a busy highway.

lilbean | October 26, 2016

I had many false positives but emergency braking never kicked in. I had AP slam on the brakes when it read the truck speed limit. If there had been a car behind me, it would have rear ended me.

kevin | October 26, 2016

Which firmware version?

Silver2K | October 26, 2016

they can pull camera recordings according to a report about a month ago. You would think they would pull the recording and use it as a reference.

Silver2K | October 26, 2016

he def. has 8.0 from that description

stuart.saunders | October 26, 2016

When I took delivery version 8 was installed.

Are events like this common? Any advice to prevent this besides disabling emergency braking?

Tropopause | October 26, 2016

Fleet Learning must take place.

GHammer | October 26, 2016

Since this is the first time I have heard of something like this and I have never had it happen to me or my wife, I would say this is very uncommon.

martin | October 26, 2016

I have had a few false positives but never on a straight road. In a tight curve the car does not correctly assign radar detected cars correctly to lanes. Service knows about this.

Tâm | October 26, 2016

The problem has been reported in the industry and not just Tesla:

http://www.wsj.com/articles/car-regulator-probes-complaints-on-brakes-14...

Some just turn the feature off instead of filing formal complaints so it looks like the problem is not major.

Tesla's blog also explained how a radar would be fooled:

"A discarded soda can on the road, with its concave bottom facing towards you can appear to be a large and dangerous obstacle, but you would definitely not want to slam on the brakes to avoid it."

Version 8 should have reduced this kind of problem but there's no guarantee.

That's why there's AP2 with more cameras, sonars and better radar.

If that still won't do it, may be adding a LIDAR might help.

In the mean time, keep formally reporting the incidents so Tesla can refine the system.

martin | October 26, 2016

Tam:
A LIDAR is a very expensive piece of very delicate optical equipment and I would not want one in a car I use every day.
And a Lidar system only works well in clear wheather. Dust, rain, fog or snowfall severely limits its usefulness.

martin | October 26, 2016

Recently I saw a Tesls Model S with two Velodyne LIDAR Systems on the roof at the Service Center. They apparently were doing autonomous driving research for Tesla. Also on the roof was a GPS receiver with an additional signal antenna increasing accuracy to a few cm in space. It was hooked up to a laptop using the maps from the "here" consortium that the major European car makers have bought from Nokia to keep it out of the hands of Google.

Tâm | October 26, 2016

@martin

Thanks for sharing your report of sighting the research car at a Service Center. Very cool!

1) If LIDAR cost is out of reach, then it's a good reason for its absence in Tesla.

2) LIDAR reportedly works with rain and snow:

http://velodynelidar.com/faq.html

"How well does Velodyne LiDAR work in snow?

Velodyne's LiDAR sensors work well in snow, sleet, and rain. The multiple beam approach of Velodyne's LiDAR sensors with laser beams with millions of laser beams at different angles enables to find "holes" in-between the snowflakes to "see" the environment. An inferior LiDAR with only one or a few laser beams would not work as well as one with 16, 32 or 64 laser beams."

Ford demonstrated its capability during January snow day:

http://qz.com/637509/driverless-cars-have-a-new-way-to-navigate-in-rain-...

No body argues to take Radar away. They only say if your existing hardware don't work in certain scenarios, why don't you add LIDAR.

If the reason is because LIDAR doesn't work in inclement weather, then let's use it in good weather because some other sensors do work in inclement weather already.

martin | October 27, 2016

Tam:
Have you seen the privets of the Velodyne Lidar systems? They're in the 5 digit range and not at the lower end.
Independent research shows that rain, dust and snow significantly reduce the signal to noise ratio (decreasing signal and increasing noise) thus requiring tremendous computing resources just to separate the good from the bad returns. Using multiple lasers helps off course (or of course?) but still.
The Google car uses a single 64 channel Lidar system delivering a maximum of oder 20 million measurements a second!

martin | October 27, 2016

Privets -> price range

Anthony J. Parisio | October 27, 2016

I have had two false positives but neither were emergency stopping. It is not common. It did not happen in the same year. Iit happened on v7 and v8. The car began to slow with no other car or object to be found. Neither event was a dangerous situation. I really don't know what the car was seeing but my gut tells me it was not a malfunction. I believe something was reflecting the radar.

dxn | October 27, 2016

Recently I encountered an event where there was a car on my left lane just ahead of me and my autopilot somehow interpreted as it was in front of me and displayed in red and engaged braking. The road was straight. I override it by stepping on the gas to prevent being rear ended.

rg22.vanhorn | October 27, 2016

I have had some false positive collision alerts while using TAAC under V8.0, though very rarely. I'm sure not going to stop using TAAC / Autopilot just because of this. It's better for Tesla if we continue to appropriately use these systems so they continue to gather data for better Autonomous driving in the future.

PhillyGal | October 27, 2016

I don't think it's very common either. I've had a few false positives on collision warnings but never auto braking while using AP. I'm glad there were no ill effects of your experience.

Run4Waffles | October 27, 2016

I've have had 3 false positives and it has made me turn my head sideways like my GSD. Never in the same spot or the same conditions. It made me a bit apprehensive when it happened but I still use AP. It's not going to be perfect and no one was behind us thankfully. Always vigilant and responsible.

Tâm | October 27, 2016

@martin

1) I have never heard from Tesla that it doesn't use LIDAR because of cost.

Tesla's reasons for the absence have been:

1) "unnecessary"
2) Radar works in inclement weather

I am no engineer so I can't disprove that "unnecessary."

However, I would not understand the rationale that just because LIDAR only works in good weather then we should not use it.

So far, Autopilot related accidents that owners have reported so far (some confirmed by Tesla, others dismissed by it) have happened in good weather with clear visibility.

None of those accidents happened in fog, rain, sleet, snow....

There's no denial that LIDAR is impaired in rain, sleet, snow. As stated on Velodyne's website, that's why to compensate for that, it recommends to use more number of laser beams rather than less.

TeslaTap.com | October 27, 2016

I'm indifferent on LIDAR, but so far Tesla has shown it is not necessary. There are many ways to accomplish the goal and some may use LIDAR and others radar/video. The car in the service bay with LIDAR may be the Stanford University car, which they are doing autonomous testing. I don't think it has anything to do with Tesla or future Tesla's designs. They bought a Tesla and added LIDAR on top of a custom roof rack.

Tâm | October 27, 2016

@TeslaTap.com

There have been speculations that LIDAR would prevent a fatal Autopilot accident but there's no proof until someone can setup a recreation of an accident scenario but this time, with an addition of LIDAR.

http://www.valuewalk.com/2016/10/tesla-motors-autopilot-big-flaw/?all=1

A model S with LIDAR and MFG for Manufacturer plate:

http://www.teslarati.com/another-tesla-model-s-spotted-lidar-test/

TeslaTap.com | October 27, 2016

@Tâm - Good catch - I wasn't aware of that one. Looks like the road is Page Mill in Palo Alto (the building in the background looks familiar). If so, it's about a mile from the Tesla Headquarters.

martin | October 28, 2016

So there seems to be quite a bit of testing Lidar. I have seen in Germany as well. The relationship to Tesla is not always clear but my service manager let slip that the car one germany was doing contract research...

martin | October 28, 2016

But a practical implementation would require a solid state solution (no moving mirrors inside) and at a cost comparable to Radar.

gguinto | October 28, 2016

I too have experience a false positive while backing into my garage. I was backing very slowly and the wall is about 8-10 ft away when the car suddenly stopped. I thought I hit something but there was nothing behind me but a wall 8-10 ft away.

That got me thinking, with all these Tesla crashes stories, why did the emergency braking get triggered in those situations?

DonS | October 28, 2016

Cheaper, smaller LIDAR is coming from multiple manufacturers. LIDAR's advantage over RADAR is much better resolution due to the wavelength of the signal. Tesla did not omit LIDAR because RADAR is better. They omitted it because cheap LIDAR is not available and Tesla found a "good enough" solution using RADAR.
Note that all these technologies sample at high speed and make decisions over multiple samples. Algorithms that use too few samples run the risk of a couple of well placed insects being perceived as a roadway obstacle. In engineering terms, using multiple samples improves the signal to noise ratio.

lilbean | October 28, 2016

@gguinto, My car easily puts itself into P if I lift my bottom up even a little bit while in reverse.

Run4Waffles | October 28, 2016

@gguinto - I believe @lilbean has it correct. You lifted your butt off the seat as you were backing up. IIRC emergency braking doesn't work in reverse.

lilbean | October 28, 2016

I wish the butt lift "feature" was only active during AP and auto park. I'm short and I really need to be practically standing up while in reverse.

Run4Waffles | October 28, 2016

@lil - you have a 17" screen and side view mirrors. Pretend it's a truck and practice. ;-p

I know, I hate using it too. LOL

lilbean | October 28, 2016

@waffles, Haha! I've tried everything. Maybe I need to put a weight on my seat while backing up. Lol.

djmichaelmayhem | October 28, 2016

This happens from time to time. Pay attention to the road is all I can say. I had a fox cross the path of my car and it tripped, once a plastic bag floating on the freeway. its not a perfect tech, it's damn good, but not perfect.

gguinto | October 28, 2016

Thanks lil and r4w, that must've been it. :)

lilbean | October 28, 2016

:-)

mark.lawler | October 29, 2016

I had two false positives this week on the freeway at night, but attributed them to large leaves blowing about. The second time the leaf blew onto the windshield and stuck there for about a second before the wiper dusted it away.

MilesMD88 | October 29, 2016

welcome to Beta Testing!