Tesla Model S Collision Avoidance With Human Test Subject – Videos

JUN 26 2016 BY JAY COLE 22

KmanAuto and Mike Anthony recently got together at Kman’s annual Tesla & EV Get Together BBQ, and decided to see how summon, AutoPilot and and TACC would react to an actual human in its course in a live test.

Model S Summon Faces Human Interference

Model S Summon Faces Human Interference

The above video gives perspective from the driver’s seat via KmanAuto’s Youtube channel, while below is Mike’s perspective as a pedestrian via his Mike Anthony Autosports YouTube channel

Editor’s Note:  While this seem obvious to state, in no way should anyone attempt this on their own.  Many of the features tested in this videos are in Beta/intended for highway use only at this time…and regardless of that, they certainly were not intended to deliberately be put to the test.

The participants in these videos took extraordinary precautions to avoid hard, and also assumed all risk on themselves.

Testing of summon was put to various tests:

  • direct/center block of the car
  • front quarter block of the car
  • “walk in front” of car while in operation

Collision warning and avoidance was also tested with Guinea Pig Mike stepping out in front of the Model while it traveled at 18 mph in both Autopilot and TACC (Traffic Aware Cruise Control) modes.

In the end the Model S performed as intended in summon, and did alert the driver to a danger in Traffic Aware scenarios.  Autopilot?  Well, not exactly put to a test it was designed for.

Categories: Tesla

Tags:

Leave a Reply

22 Comments on "Tesla Model S Collision Avoidance With Human Test Subject – Videos"

newest oldest most voted
European point of view

Not impressive .
Remember the collision avoidance with “objects” of the Model X falcon wings .
It works and sometimes not .Refer to the most recent update of the software .

I am more interested in a similar test with a soft inanimate object. Because it could be that the Tesla had enough time to stop, but the driver didn’t want to take the risk. The Tesla will first make a sound to warn the driver and then intervene. But with this video, the driver intervened at the same time as the sound. At least not at the same time as the Tesla would have braked.

My thoughts exactly. I don’t know the behavior but it seems plausible it may be designed to alert the driver first, then take action if the driver fails to respond.

SparkFiatOwner-M3reserved

No real emergency braking at all. Subaru and mainstream companies seem to have it in standard production without beta disclaimers.

Sure, not AP, but safety over sexy — SMH at Tesla for not get safety first.

I’m wondering the same thing. How is it the Nissan is offering automatic collision avoidance even in their Sentra? I am thinking the new LEAF is going to surprise us with some remarkable technology.

Tesla no pedestrian detection, thats a joke? But KmanAuto should let do the tests NCAP:

Thanks for posting rational, calculated and safe testing, as opposed to cowboys looking to become a statistic.

I’m sure this is one of the areas where Tesla are working hard. It’s a young company and safety takes time to develope.

They have been doing surprisingly good in many areas, but in some like safety they have a lot of catching up to do.
It could be a lot worse though…

Why not just put a piece of foam in the road vs. having a person jump out?

Did you forget what happened when Bjørn Nyland tried that with a “wall” of styrofoam?

I’m not sure, but my guess is that the long wavelengths of radar don’t bounce off something as low density as styrofoam; that radar goes right thru it instead of bouncing back.

http://insideevs.com/bjorn-nyland-tests-tesla-model-s-automatic-braking-feature-video/

Yeah, it was the same result. Tesla’s collision avoidance didn’t stop for the styrofoam wall in Bjorn’s video, just like it didn’t stop for the human being in Kman’s video. Are you saying that radar goes right thru a human being instead of bouncing back?

He is just saying styrofoam doesn’t work for a human analog. The rest is your own extension.

“Are you saying that radar goes right thru a human being instead of bouncing back?”

I hope you don’t seriously think that’s what I was saying. Or even remotely close to it.

I really hope you’re kidding here, sven.

* * * * *

The way to do test this properly would be the way that TV’s “Myth Busters” do it. Use a mannequin made of a gel that has the same density as human flesh, with something resembling a human skeleton inside. That should approximate the radar signature of a human being.

Put water bottle in the trousers and it works. Also you might need a human like face + clothes for the camera.

“Editor’s Note: While this seem obvious to state, in no way should anyone attempt this on their own.”

It seems like it should be obvious that nobody should try to act as a human crash test dummy.

Sadly, it seems that common sense gets less common with each passing decade.

This is insanity, testing with live humans in front of a 4500 pound car.

Yes, thank you for putting your lives on the line instead of us. But please don’t do that again.

Even Daimler Trucks can handle pedestrian in 2016, why can’t tesla when their camera (from mobileye) can already detect humans pretty well? Or did tesla change the camera producer?

A extensive test with the 4a test dummy would be interesting.

So far as I know, nobody developing a self-driving car uses cameras to detect obstacles. The cameras in cars are to help people see out the back and in blind spots.

Google’s self-driving cars use a rotating lidar scanner. Tesla uses non-rotating radars and ultrasonic detectors.

Speaking as a computer programmer: Computers and software are lousy at recognizing things in camera images. It’s true that facial recognition software does a surprisingly good job, but that’s a special case, and very different from the ability to detect a human of any size and shape, wearing any clothing, in any posture, from any angle.

scott franco (No M3 FAUX GRILL!)

True, but computers do better with dual stereo cameras at sorting out approaching objects, using parallax.

I personally think lidar makes the most sense for cars, because of it’s accuracy and deterministic behavior.

Talk of days of which they sit and wait, all shall be evs…
Good song choice K-man.
https://www.youtube.com/watch?v=c9P34hJ0WEY

scott franco (No M3 FAUX GRILL!)

“you did WHAT with the baby in the car? You are in such trouble…”