from the there-is-a-conference-for-everything dept.
This Anonymous Coward is signed up for a variety of trade magazines and shows relating to the auto industry. Here's the announcement for a new conference that just came through: https://autonomousvehiclesymposium.com/detroit/en/. It notes that the Autonomous Vehicle Test & Development conference is next month in the Detroit area:
Advanced driver assistance systems [ADAS] giving rise to fully automated driving vehicle technology is nothing new. Since the final meeting of the Eureka PROMETHEUS Project in Paris in 1994, it's been clear that fully autonomous self-driving vehicles are set to become a reality, yet 24 years later the final stages of testing, validation and fail-safing pose a huge challenge to the automotive industry.
The speakers are listed with a blurb, here are a few I found interesting:
What to do about the next 500 fatal AV [Automated Vehicle] crashes: The presentation discusses and analyzes the current status of AV verification, using recent examples. It analyzes challenges to eventual deployment, noting that we can expect many fatal AV accidents. It then suggests that a comprehensive, transparent verification system could help solve this inevitable tension. Finally, it describes principles of verification using a scenario-based, coverage-driven methodology.
Toward Vast Scale Virtual Validation - High Fidelity Road Database for Simulations: ADAS Simulations will allow to simulate autonomous driving of millions of miles per day. This is largely considered a prerequisite to validate autonomous vehicles and simulation vendors tailor their software to that end. Realistic content for simulations, however, is cumbersome to source and no database of real scenarios is available. We will present how to efficiently source 3D map data and real driving scenarios for use in simulations and introduce our database of thousands of kilometers road length. This database is captured in different automotive hot spots and contains centimeter-accurate digital twins of road networks. These can be used as content in simulations.
ADAS testing advanced: 6D target mover: Pedestrians and cyclists account for a significant proportion of road deaths worldwide. Current ADAS test systems are tackling this challenge, but are limited in their design to linear or two-dimensional motion. With this setup, particularly during acceleration processes, an unrealistic motion is generated. The concept of hanging dummies from above creates new possibilities for more life-like dummy trajectories using six degrees of freedom. The system sets new standards in precision and repeatability through the ability to reproduce real-life human motion sequences and imitate them realistically – for example, based on data from a motion capture system.
So, tell me again, how is it that a teenager can be given some limited instruction and (in most cases) drive for a lifetime pretty successfully. Seems like there is a large gap between "AI" and "I".
(Score: 0) by Anonymous Coward on Thursday September 13 2018, @12:26AM
a teenager is artificially stupid?
(Score: 3, Touché) by jimbrooking on Thursday September 13 2018, @12:41AM (1 child)
As the AI industry evolves more idiot-proof systems, human evolution evolves smarter idiots.
(Score: 2) by Snotnose on Thursday September 13 2018, @12:46AM
Doesn't work like that. I think the most famous pot smoker in history said something like "It's gonna explode. But it's gonna explode in a different way".
/ drunk me fighting back
When the dust settled America realized it was saved by a porn star.
(Score: 2) by takyon on Thursday September 13 2018, @12:48AM (4 children)
Were you under the impression that autonomous vehicles would cause no deaths? Some accidents are impossible to avoid.
If AVs cause 15-20% of the deaths that human-driven vehicles do, then of course there's going to be a "next 500 fatal crashes".
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 1, Insightful) by Anonymous Coward on Thursday September 13 2018, @02:27AM (3 children)
If they weren't being driven only in easy conditions, and were being driven more, I'd be less concerned. As it stands now, the cars fail miserable with easy conditions leading to cars driving into off ramps, over motorcyclists, into fire trucks and the like.
The vehicles are nowhere near good enough to drive on public streets while they can't figure out how to handle those things reliably. Drivers who aren't under the influence of drugs can easily handle all of those things without trouble.
(Score: 2, Disagree) by MostCynical on Thursday September 13 2018, @04:49AM (2 children)
Drivers, with or without distractions, do a lousy job of not hitting things. In fact, humans are shit at it.
http://www.who.int/news-room/fact-sheets/detail/road-traffic-injuries [who.int]
"I guess once you start doubting, there's no end to it." -Batou, Ghost in the Shell: Stand Alone Complex
(Score: 1, Insightful) by Anonymous Coward on Thursday September 13 2018, @06:25AM (1 child)
Seems more like males age 25 and under are bad at driving, and even they're not too bad. Out of hundreds of millions (at least) of people driving a day, only a few thousand die?
(Score: 2) by MostCynical on Thursday September 13 2018, @07:11AM
If only people managed to just kill themselves, not others, and pedestrians and cyclists..
"I guess once you start doubting, there's no end to it." -Batou, Ghost in the Shell: Stand Alone Complex
(Score: 0) by Anonymous Coward on Thursday September 13 2018, @03:53AM (4 children)
Why do airlines have such strict and onerous conditions and certification processes before they are allowed to fly, but apparently anyone who wants to is allowed to build and sell autonomous or semi-autonomous road vehicles?
(Score: 1) by khallow on Thursday September 13 2018, @04:14AM (3 children)
Because when a passenger airplane fails, it kills many dozens or hundreds of people and generates an impressive media frenzy. The first death by autonomous car just didn't have that media draw.
(Score: 0) by Anonymous Coward on Thursday September 13 2018, @07:56AM (2 children)
That doesn't make sense.
(Score: 2) by acid andy on Thursday September 13 2018, @12:10PM
I think humans might have an irrational subconscious belief that an air crash should be slightly more frightening than a vehicle crash, even if both are potentially fatal. I suppose on reflection it's not all that irrational. Motion perpendicular to the ground (due to gravity) is inherently more dangerous than motion parallel to the ground. Sure, parallel to the ground, you might wind up hitting a small object sticking up from it, but perpendicular to the ground, the ground is a huge object in your way, much harder to miss.
There's also probably the aspect that as a helpless passenger, you have no control over the plane, but if you're driving, you have some control over the vehicle's, and therefore your own, destiny. Even as a backseat driver, you have more influence than an air passenger. Autonomous vehicles obliterate that advantage though of course.
If a cat has kittens, does a rat have rittens, a bat bittens and a mat mittens?
(Score: 1) by khallow on Thursday September 13 2018, @12:39PM
First, is my assertion correct? Was indeed the first death by autonomous vehicle more significant than the latest passenger airline crash with a lot of fatalities? I think it was. In the end, one person's death by a mundane car accident (even when Uber was involved) just doesn't bleed enough for most media sources.
Second, why would I speak of media attention? Here, we need to look at this as regulators and elected politicians would look at the situation - "What do I need to do to cover my ass?" The big media events are a big threat to their careers while small ones aren't. A lot of regulation and its enforcement is about minimizing the risks to the regulators, not doing their job well.
I believe that's why airliner activity is far more heavily regulated than experimental autonomous vehicle operation.
(Score: 1, Informative) by Anonymous Coward on Thursday September 13 2018, @04:26AM
As I understand it, one of the big problems is figuring out why an AV control system made a particular decision. Looks like some researchers are working on that problem, this might be one?
https://www.researchgate.net/publication/325543751_Explaining_Explanations_An_Approach_to_Evaluating_Interpretability_of_Machine_Learning [researchgate.net]
(Score: 2) by bootsy on Thursday September 13 2018, @08:03AM (7 children)
No one has yet been able to explain to me how AI cars will handle situations where you need to judge what another driver is going to do. I've been driving down a lot of country lanes during the summer in various different parts of the UK and they are often only the width of 1 vehicle with pull in points. There is a frequent need to reverse back to a pull in point and it is usually the smaller vehicle that is expected to do this. Very often I will flash my headlights at someone to indicate I am happy to stay back and let them through.
When pulling out at roundabouts and junctions or spotting someone about to pull out, it can often be helpful to look at the head of the person in the car you are worried about. If they aren't looking then you can almost predict they will pull out recklessly and so I can take action to avoid that, i.e. slow down, go wider and potentially beep the horn.
How can autonomous driving systems know what other cars will do unless they have some form of communication between them and then what will they do when they come across a non-autonomous vehicle?
If I see that someone is poor at doing hill starts then I will leave myself even more space between us when I come to a stop on a gradient. If the car in front and the car in front of that and the car in front of that are all bumper to bumper on a motorway I will leave even more of a gap as I know if one of them hits the brakes they will all pile into each other.
I was driving in the summer and I saw a stall selling strawberries in a layby. I slowed down as I knew drivers would be distracted looking at it and some may decide to stop suddenly and buy them. I am glad I did as the three cars in front of me promptly crashed into each other. Even if an AI could detect the sudden stop is it would not be distracted, the sudden breaking could cause the car behind to crash especially if it is a big lorry that is due a service.
If you have a roundabout with 3 exits and a car arrives at each entry simulataneously then how does the AI decide who should go first?
(Score: 3, Funny) by c0lo on Thursday September 13 2018, @09:56AM (4 children)
Here yea go.
AI handling other driver procedure:
- step 1. If other driver is human, kill I
- step 2. If other driver is AI, establish a network connection and share expected maneuvers in advance
Simple. In a few years, there'll be no other drivers than AI, problem solved.
https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
(Score: 2) by bootsy on Thursday September 13 2018, @12:54PM (3 children)
You been modded funny but I have feeling your prediction will be accurate and that it was become increasingly difficult to be allowed to manually drive a vehicle on the public roads.
(Score: 2) by c0lo on Thursday September 13 2018, @12:57PM
Yeah, well, what can one do? Laugh while you still can.
https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
(Score: 0) by Anonymous Coward on Thursday September 13 2018, @02:30PM
I'm early 60s and I'm not worried. Where I go (mostly all 'burbs and country roads), I'm convinced there will be many of us old school normal, manual drivers for as long as I want to drive. Maybe not so many using stick shift, but I enjoy that too. Since my father was still driving fairly well in his late 90s, I figure I've got another 30+ years of driving enjoyment left in my life.
Re eye contact: I saw something recently where a test AV was fitted with large animated cartoon-character eyes, the idea was to signal to other drivers and peds/cyclists that the car had "seen" them. Maybe the AVs will all wind up with something like this on the front?
(Score: 2) by Knowledge Troll on Thursday September 13 2018, @02:34PM
Some people I talk to don't even realize that they are advocating for taking away steering wheels from humans but they are. I've had a few conversations like this:
me> I don't like autonomous cars
them> But they'll reduce human fatalities to zero!
me> I'm not so sure about that and I'm worried they are going to make it so humans can't drive
them> whoa whoa whoa no one is talking about taking away your steering wheel man, stop being so paranoid
But ok - how are you going to stop nearly all deaths by replacing people with robots with out replacing people with robots?
(Score: 2) by Knowledge Troll on Thursday September 13 2018, @02:39PM (1 child)
I think that the conflict between laws on the books regarding driving and roads and the reality of driving in the world is going to be pretty strong. I can't fathom any robot car will do anything besides exactly the letter of the law because anything else just opens them up to legal risk. But driving in a city or lots of other situations means you skirt the rules just to keep traffic flowing. Humans will do this but the robots won't.
The way that they'll handle the situations you describe is by following the letter of the law to the T.. How is it going to play out? Some driver has a problem starting on hills, robot car stops right behind them treating the situation as every other one, the moron in the front accidentally rolls into the robot car you are sitting in, now you get to wait for the police and insurance company from the robot's handlers so you can give a report; hope you didn't have anywhere important to go!
(Score: 0) by Anonymous Coward on Thursday September 13 2018, @02:54PM
> Some driver has a problem starting on hills,
Guessing you are not in USA? Here the automatic transmission has taken over nearly completely, and none of them have problems getting going on hills. Those few of us that still drive stick shift mostly do it out of enjoyment and know how to either coordinate pedals, or use the parking brake to start on hills with little drama.