Tesla Autopilot Tested In The Snow – Video

NOV 2 2015 BY MARK KANE 24

Tesla Autopilot Tested In The Snow

Tesla Autopilot Tested In The Snow

One of the Iceland’s Tesla Model S 85D owners published a very interesting Autopilot test, just after he downloaded the update. But at the same time, Iceland downloaded some snow.

It’s especially interesting because unlike many other videos when everything works fine, this time the Tesla had to deal with snow.

Autopilot can drive the car when there is no snow and to some degree even when part of the road is covered slightly, but when the conditions got worse, the Tesla isn’t able to drive itself without losing the track and turning off the Autosteer.

The author of the video does not encourage the use of Autopilot when the road is covered by snow. Thankfully, Iceland’s winter is just 11 month long… 6?

“We got the autopilot update last night. When we woke up this morning and wanted to do a test drive the fist snow of the winter had fallen.”

Categories: Tesla, Videos

Tags: , ,

Leave a Reply

24 Comments on "Tesla Autopilot Tested In The Snow – Video"

newest oldest most voted

Gawd i hate snow!

Some iterations later, or some improvements in the infrastructure (emitters in the road perhaps) will allow autonomous travel even in snowy conditions.

As to snow, I couldn’t live anywhere where there is no winter. It is my favourite season. I’d rather have snow than humid, sticky, dirty summer air πŸ™‚
Let it snow.

+10

Darn! I hate that auto-veer!

Good one.

Me too! I live in south Florida.

Data collected from Tesla Vehicles when the roads are clear, will help resolve that issue.

The problem with the data collected is the data’s inertia as the roads change (i.e. due to construction and closures) and the accuracy of the vehicle positioning to follow the such data. A few feet to the left or the right may result in a headend collision or a leap from a ravine.

Lots of rainy day scenarios to identify and resolve…

The vehicle also monitors temp and traction data… The issue is not unsolvable.

Logic dictates that in poor conditions use of the autopilot should be reduced or not used.
In snow, dust storms, fog, high winds, and other adverse conditions it should not be used.
In fog I wonder if it might help, in that infrared detectors could help avoid some of those infamous pile-ups on the Pacific Coast Highway, that resulted from foggy conditions.

Highway Safety:
“Despite the relatively low prevalence of fog- and smoke-related fatal and police-reported
crashes, and a general decreasing trend in such crashes over the past two decades, fog and
smoke remain significant threats to highway safety given the particularly insidious ways in
which these conditions appear to impact driver perceptions and behaviors. Given the
increased likelihood of crashes in the presence of fog and smoke – and, most troubling, the
increase in severe and multi-vehicle crashes – fog and smoke should be treated as serious
safety concerns, and efforts should be made to continue developing and evaluating
countermeasures targeting the issue. The national data highlight certain priority concerns
with regards to fog – such as young driver crashes, winter driving, multi-vehicle pileups,
and undivided rural roadways – and, when considered alongside existing research into
driver behavior and perception, offer actionable ideas for highway officials, safety advocates, parents, automakers, driving instructors, and other stakeholders wishing to see reductions in accidents.”

BMW LASER lights are a huge improvement, when it’s snowing.

Good to know. I suppose a less diffuse light source accounts for that.

Any conditions that are iffy, it would be prudent to use human senses.

I did use ELR’s ACC in fog this morning. First time in fog. Did pretty well. Everyone was going so slow that I wasn’t overdriving visibility anyway.

And what is this ACC you speak of? Care to elaborate?

From a previous post by Loboc:
“My ELR has ACC (Adaptive Cruise Control) as do many other cars out there (some 10 years old!). The car is aligned, so, it tracks straight on a straight road as demo’d. Very little driver input needed to keep in lane and the go+stop pedals are automated. It also has lane departure alert.”

Thnks! Should have known this.

Yeah the car is learning but it doesn’t need to see the lines on the road. When the car can’t see the lines it will use high resolution GPS but if that’s the first Tesla on those roads the GPS data is not available. Try a month from now and see how the car does then.

Civilian GPS isn’t accurate enough, it can be off by 15 to 30 feet.

http://www.gps.gov/systems/gps/performance/accuracy/

Interesting challenge for the future, for self-driving cars. I wonder what the solution will be? Perhaps, as Ontario Leaf suggested, it will require wireless emitters in the roadway.

@Darin:

GPS does not provide an accurate enough location for lane-keeping. Heck, even with military grade GPS, the smart weapons have a two-foot margin of error. And the civilian GPS has been deliberately downgraded from that.

Perhaps in the future, self-driving cars will use terrain scanning radar combined with an internal digital terrain map, such as is found in the guidance system of cruise missiles.

If humans don’t need a terrain map, neither do cars. Their vision systems just need to learn about snow.

Humans do a bang up job driving. πŸ™‚

I was always taught to not even use regular cruise control in bad weather. You need to be scanning ahead for black ice, deep water, etc.

amen – 70’s RWD boat, cruising interstate in poor weather (no precipitation), jamming to tunes.

Bridge had typical bump-joint, slowed car slightly, transmission-sensor cruise was unpleasantly unaware as acceleration was applied and continued and rear wheels broke loose inducing an instant 30 degree offset on invisible bridge ice @ 60mph. Not a fun lesson (first ‘nice’ car with cruise control).