Why Autonomous Vehicle Safety Statistics Need Standardization

As companies make ever-impressive strides in vehicle safety and sensor technology, we're seeing more and more semi-autonomous vehicles hit the roads.

One argument used for their adoption is safety statistics which clearly show semi- and fully-autonomous vehicles are (usually) safer than human drivers.

But as one academic study uncovered recently, there's some significant issues with citing safety statistics of vehicles with semi-autonomous features when their use is restricted to only some types of road - roads which traditionally have fewer accidents.

Here's our deep-dive into those studies - and what the implications are for autonomous vehicles moving forward (even if they're still safer than humans alone).

---------
00:00 - Introduction
01:32 - Subscribe and support the show!
01:50 - More companies than ever before using semi-autonomous and ADAS systems
02:13 - Tesla’s safety statistics V Regulation
03:33 - The challenges of non-standardized safety statistics
04:15 - What is ACTUALLY considered an "accident"?
05:09 - Autonomous and partially autonomous more likely to be hit
06:34 - Some points to bear in mind based on current fleet technology
08:50 - Autopilot STILL safer than non-autopilot *with a smaller margin
11:24 - Thanks, and goodbye!

----------
Presenter, Camera, Script, Coloring : Kate Walton-Elliott
End Credits, Artwork: Erin Carlie
Audio, Editor, Producer: Nikki Gordon-Bloomfield

© Transport Evolved LLC, 2022

This video was made possible thanks to the kind donations of our supporters on Patreon and Ko-fi. Join them in supporting us.

Join this channel to get access to perks:

Buy the team a coffee -- or a meal!

Donate using bitcoin: BC1QRZRGCCXUKPDZ2DZ9KX9JYJXMZDYHYGFZ0MN4WH

-~-~~-~~~-~~-~-
Please watch: "2023 Kia Niro EV: Why You'll Want To Drive This"
-~-~~-~~~-~~-~-

[jestempies]

I really like that you've taken to publish more editorial-type videos. News is fine and reviews are great, but a good editorial beats both :)

[senordockman]

My personal experience using FSD on two lane roads at speeds of 45 mph or less has been very good. In fact, I now engage FSD as often as possible in traffic because it usually is very relaxing in stop and go traffic and will keep me out of trouble. I do however disengage FSD on certain situations in the city. Anecdotal evidence and personal experience indicates to me that autopilot is safer than my driving in most driving situations that I have experienced so far after 10, 000 miles of use.

[beachcrow]

So here's Tesla's Q4 2021 report: _In the 4th quarter, we recorded one crash for every 4.31 million miles driven in which drivers were using Autopilot technology (Autosteer and active safety features). For drivers who were not using Autopilot technology (no Autosteer and active safety features), we recorded one crash for every 1.59 million miles driven. By comparison, NHTSA’s most recent data shows that in the United States there is an automobile crash every 484, 000 miles._ Note the Tesla stats are comparing only between Tesla vehicles, not the general population. Any accident safety data can get confusing fast as though there are more crashes in urban areas vs limited access highways, the rural non-divided highways have a far greater number of serious crashes then slow speed urban crashes. Are we counting injured people or banged up cars?

[williammeek]

The problem is that Tesla doesn’t have access to the data to make a more apt comparison and those agencies that do are actively avoiding publishing such data.

Sure that data isn’t a good match, but it IS the best information we have.

[AnonymousFreakYT]

Love the NASA shirt with the "monolith" CPU ear rings! (What are those, Motorola 68000s?)

[jom]

Thank you for that very informative video !! I'm still not sure if I like the idea of my car driving itself. But driving aids are always welcome.

[johneric]

Good reporting! I had wondered if the comparisons were like-to-like. Apparently not. Also, those owning Teslas tend to be older thus generally safer drivers.

[dimasalves]

I love that ya'll <3 Star Trek as much as I do. :)

[tommays]

Your really on a strong journalism Roll

[robingoergner]

Kate, your reporting is spot on. Any assertion that autonomous driving systems are safer have to backed up by objective data and careful analysis. At this time, I don't think we're there yet.

[SteveRowe]

Insightful as always, Kate. I love your technical approach to topics and your easygoing presentation style. I agree that we need more details for each incident, but it will always be far to easy to mislead with statistics. People just don't want to do the homework.

[CaptainXJ]

We need 100% auto pilot with an industry standard.

[olebloom]

Balanced and fair reporting as always.

[theq]

I always dismissed that comparison because on average tesla cars are new and newer cars are less likely to have accidents than older ones on a mechanical point of view, even if we ignore all now standard features the probability of something failing and causing an accident increases with time.

[niceracleous]

When a child is learning to walk, they fall over a lot till they’ve mastered it, this is the same thing with autonomous driving at this point in time, it needs improving but it will come

[TheJAMF]

5:20 That number should now go up for Tesla, when they stop at a stop-sign in a region, where a rolling stop is common. 🤔

[chrisheath]

There are so many variables in driving. But the problem may be that people simply have no idea of the risks involved. In the US there are over 17, 000 car accidents per day. And over 20, 000 deaths annually. This tends not to be made public, because probably, car manufacturers spend millions on saying how wonderful their car is, and would oppose a negative narrative. However, the high daily accident (collision?) rate indicates that human inattention is one of the highest causes, and risk and accident awareness, via public campaigns, may improve the problem.

[EntropicRemnants]

Very good. Hadn't even thought about how these statistics are compared. Thanks.

[trevorbromley-palmer]

I am so disappointed with my 2021 Tesla M3LR automatic driving that I now don’t bother to turn it on. Far too many phantom breaking events at 70Mph because I drove under a bridge, or a lorry was in the slow lane but my M3 decided it was about to crash into it by overtaking that it slammed on the breaks. Then there is the Navigate on Autopilot prompt to change lanes to overtake, but it completely ignores the fact that there is a car in the outside late directly next to me, so overtaking is not possible. Then there is the frankly alarming front collision detection that you are about to drive straight into something where the road is perfectly clear. All of these things made driving on AP or NOA very stressful. This is a shame as I was so excited to buy a Tesla and to have these features, now I just don’t think they work in the UK well enough, and would not spend the money again. My previous Kona Electric drove much better on motorways without ANY of these issues. The roads/traffic must be so different in the US, but in the UK I don’t think it is close to being safer than a human driver. Such a shame.

[kensmith]

With the safety systems in the car, the drivers will pay less attention even when the safety system has a problem it can't solve. This could mean that the overall danger is lower but the numbers for the "tricky situation" numbers for the autopilot equipped cars will be worse.
Stability control is only really needed on a car that wants to oversteer and then roll. Suspension design and limiting how high the center of gravity is can make a car that tends to understeer a little and remain shiny side up with no software needed.