In a stunning example of the Baader Meinhof effect, we’ve recently heard several times this week about events like the “carbage run.” That is, a motoring event where you can only buy some garbage car to compete. In the case of [Robbe Derks], the idea was to take a six-day journey to the polar circle in a car. But not just any car. It had to be at least 20 years old and cost less than €1000. That wasn’t hard enough for [Robbe] and friends. They also decided to make the car self-driving.
If you have a car that is new enough, this might not be as hard as it sounds. The OpenPilot project adds L2 self-driving features to about 275 car models. But probably not a 20-year-old junker or, in particular, a 1993 Volvo 940. [Robbe] took up the challenge and is doing a series of blog posts covering how it all worked.
Most (or maybe all) cars in 1993 didn’t have actuators for remote steering, so the car needed a transplant from a 2020 Toyota Corolla part. Adaptive cruise control also needed some help in the brake system. Add an accelerator servo and an optional radar sensor and you are almost ready to go.
We are waiting for more blog posts to tell us just how close to ready you are at that point. But even the first post has a lot of cool car info. It won’t be a weekend project to duplicate, but it does have a certain cool factor.
Now add a decidedly non-1993 Android phone… If you want to start with something less complex, maybe settle for driving assistance only in certain conditions.
Self driving cars were achieved in the 1980’s. The question is merely about the definition – whether it’s succeeding 95% of the time, or 99% of the time, or 99.9%… etc. How often does the driver have to intervene?
A person might drive a million kilometers in their lifetime. How many accidents they’ll have on average during those miles? That sets the standard for how good the computer has to be, to be worth it for the average driver. That however is a biased estimate, since the average person includes all people who drive drunk or on drugs, in junk cars, on terrible roads and conditions, etc.
well there was a radio controlled car in 1925
page 26
https://www.worldradiohistory.com/Archive-Radio-News/20s/Radio-News-1925-11-R.pdf
and others in 1950’s
Nearly everyone thinks they are better than average.
The self-driving cars? Pretty sure that’s wrong if it’s measured even halfway honestly
I meant humans, so even if the self driving cars are statistically better than average in terms of accidents.
Most people will think they are better than average themselves.
This is almost completely irrelevant to the discussion of whether self driving cars are “good enough”. What people look at are a perception of accident rates and general safety, not an actual comparison to human capability. Fire example, airline crashes are seen as terrifying enough to prevent some from flying away all, yet are statistically much safer than most simply because of the immediate death toll.
At any rate those vehicle is separate from that discussion, being a DIY project intended for use on relatively low traffic roads at low speeds. Try driving to the arctic circle, it’s more about fatigue and physical road condition than metropolitan driving.
Given the conditions most self-driving miles are done under it’s self-selecting only the easy cases – you see the stat about how many millions of miles Teslas have clocked up but ignore the fact those are going to be mainly easy highway miles that clock up fast and pose no great challenge; There was a Mercedes van full of computers in the 1980’s that could drive on the highway pretty successfully, how much further are we from that other than making it cheaper and much much smaller?
I’ve had 3 or 4 hire cars now with modern driver assist features and most of them were wrong more often than they were right. If they were a human you’d have at least suspected they were not paying attention or inebriated half the time.
And most of them would be correct. This is because driving skills and other factors that affect the relevant metrics are not normally distributed – a minority of drivers are truly terrible while the majority are ok, while the accodents caused by the few are counted on the majority by averaging it out.
If you make the argument about the statistically average driver, you would also be making the hidden assumption that e.g. every driver drives some portion of their total miles drunk or on drugs, while in reality most will never do that.
For instance, when was the last time you drove a stolen vehicle? If you answer “never”, chances are you’re less accident prone than the average driver.
In other words, the average driver (a sample of randomly picked drivers) is likely to perform better than the average driver (all accidents per all drivers).
Paradox? No, just a common propaganda trick base on the logical fallacy of equivocation: calling different things by the same name and hoping that nobody notices. One side is interested in the probability that the next car down the road isn’t going to crash into you, while the other side is interested in the overall accident rate regardless of who is the cause or the victim.
Baader Meinhof
Baader Meinhof Effect https://en.wikipedia.org/wiki/Frequency_illusion
I’ve been hearing about this more and more recently
I see what you did there!
You win the internet today
I’m surprised my spell checker didn’t catch it given that it wanted to change carbage into carnage repeatedly.
It’s a Volvo 940. The self driving only has to be vaguely right as it’ll happily drive through anything in it’s way!
Knowing the roads in question, it’s not uncommon for human drivers to hurl themselves off of these ones dude to fatigue and poor road surface. The Volvo isn’t a bad car to take on a trip like this, but something with more clearance would be safer.
A 1993 Volvo 940 for less than €1000?
Wow, so what about this Volvo?
X4090.You can spot a second hand market autopart when they come with this white or yellow markings, I think is the same guy traveling very fast from country to country.