Simanaitis Says

On cars, old, new and future; science & technology; vintage airplanes, computer flight simulation of them; Sherlockiana; our English language; travel; and other stuff

AAAS LOOKS AT AUTONOMOUS CARS

SCIENCE, THE WEEKLY magazine of the American Association for the Advancement of Science, asks on its December 15, 2017, cover, “When Will We Get There?” Its subhead: “Deploying driverless cars might be harder than we think—and could bring unwanted consequences.”

The association’s sciencemag.org takes an even stronger stance on the matter: “Are We Going Too Fast on Driverless Cars,” by Jeffrey Mervis. His feature article in Science magazine is titled “Not So Fast,” and notes, “We can’t even agree on what autonomous vehicles are, much less than how they will affect our lives.”

Mervis quotes Senator Gary Peters, D-Michigan, who says, “This is probably the biggest thing to hit the auto industry since the first car came off the assembly line. It will not only completely revolutionize the way we get around, but [Automated Vehicles] also have the potential to save hundreds of thousands of lives each year.”

“Such predictions, however,” notes Mervis, “turn out to be based on surprisingly little research. While developers amass data on the sensors and algorithms that allow cars to drive themselves, research on the social, economic, and environmental effects of AVs is sparse.”

Illustration by Emiliano Ponzi. This and an image following from Science, December 15, 2017.

From time to time here at SimanaitisSays, I’ve identified some of these non-technical aspects of AVs, among them ethical, legal and regulatory matters. As an example, confronted with a lamentable choice of protecting pedestrians or the car’s occupants, which does the car’s algorithms choose? What are the legal implications of such a choice? And how should government regulate AV development and implementation?

A hacker’s paradise? Image from Wired
magazine, August 19, 2014.

These aspects become particularly critical with full vehicle autonomy, now characterized as Level Five of AV development. A Level Five car does everything; its occupants, operator included, are along merely for the ride. Such a travel pod would likely be devoid of any controls, other than those initiating operation, selecting destination, and shutting down upon arrival.

Six Levels of Automotive Autonomy.

The Morgan sports car I used to enjoy is Level Zero. Even its brakes, though hydraulic, were unassisted.

Level One characterizes our current fleet of cars: The driver is responsible for everything, but with aspects of automated assistance: power-assisted steering, ABS braking, often cruise control, maybe even some lane monitoring.

Level Two cars are now in transition and testing. The driver continues to have responsibility, but the car is capable of nudging itself in lane, applying the brakes in what it perceives as a panic stop, performing parallel parking, and other significant driver assistances.

Level Three gives the car full control, but only under special circumstances, on freeways, for example. The driver must be capable of taking over within 10 to 15 seconds.

Science notes that Level Three “Might never be deployed.” And, to my mind, for good reason: It’s wishful thinking that the “driver” occupied with cell phone, car display screen, or just the passing scenery could handle this 10–15-second assumption of responsibilities.

Level Four broadens the conditions under which the car assumes total control. Its only limitations would be related to speed, weather, or time of day. Notes Science, the driver does “Nothing under certain conditions, but everything at other times.”

This futuristic image would qualify as Level Four (note its steering wheel).

Alas, I see the same problem as with Level Three: It is wishful thinking indeed that an otherwise non-participating “driver” will be prepared for a patch of snowy road or sudden rain squall. Despite this, Science notes that Level Four is “Where the auto industry wants to be.”

As already noted, Level Five cars do everything everywhere at any time. Science puts this one “Somewhere over the rainbow.” Despite optimistic, albeit misleading, news reports that AVs are arriving as soon as 2020, others suggest it could be 2075 before Level Five AVs are here.

Mervis quotes Susan Shaheen, co-director of the Transportation Sustainability Research Center at the University of California, Berkeley: “The current conversation … falls into what I call the utopian and dystopian views.”

I lean toward the latter. ds

© Dennis Simanaitis, SimanaitisSays.com, 2017

5 comments on “AAAS LOOKS AT AUTONOMOUS CARS

  1. Michael Rubin
    December 30, 2017

    All I can think of is trying to get a useful result from any of the verbal command-initiated traffic services. “No, I said Sacramento, not San Mateo….” By the way, the vehicle forward and to the left in the Level 4 picture looks a lot like a 57 Cadillac.

  2. sabresoftware
    December 31, 2017

    The dictionary definition of autonomous includes: self-determining, self-governing, uncontrolled. Sounds like where we are at today. What we are looking at with automated driving is anything but autonomous, requiring either significant vehicle-to-vehicle communications or worse yet a centrally administered road network where all motions on roadways would be controlled by a big brother central computer.

    The current somewhat autonomous approach that manufacturers are proceeding with basically has each vehicle doing its own thing trying to interpret and assess external information including that from other automated vehicles, current conventional vehicles with the driver wildcard, pedestrians who can be totally unpredictable, and wildlife. Also acts of nature (flooding, mud/rock slides, downed branches, etc.) and weather.

    All the scenarios that I have seen for the ideal autonomous road system imply a purpose built roadway. The level 4 picture shows a centre-of-lane guide marking of some sort. To achieve Level 3, 4 or 5 systems that would work would require a significant investment in infrastructure that I don’t believe any nation could actually afford, certainly not in the short term.

    Achieving level 3, 4 or 5 with individual vehicles handling all the driving decisions individually would be far more complex than many people, especially the politicians and pundits who see this happening in a short period of time, realize. I have driven for 50 years now, and have seen so many interesting scenarios over the years that I strongly doubt that current and near future technology would be able to deal with everything. Hence there would still be crashes.

    And I haven’t even touched on the aspects of system reliability. I had lane proximity sensors on my old Audi Q5 that would shutdown due to sensor obstruction, when the roads were really slushy and freezing. So what would happen in an automated car? Car would make a lane change with incomplete information hoping that the other car (if any) would compensate? or the car could never make a lane change?My current SQ5 doesn’t have that system, and I don’t miss it, because once the system showed that it wasn’t up a 100 percent of the time I really couldn’t rely on it. As it was I only used it as an early warning, still relying on shoulder checks to execute lane changes.

    And to achieve a fully driverless fleet would take 10-20 years to replace the current fleet with the new technology.

    I suspect that given the amount of time necessary to fully develop these technologies to a truly reliable state combined with the fleet changeover time frame I can fairly safely say that I won’t see this in my lifetime.

    • simanaitissays
      December 31, 2017

      Given those predicting 2075 as a Level Five date, you’re not alone in this analysis. Thanks for sharing it.

  3. sabresoftware
    December 31, 2017

    In the medium term I believe that we will see systems that may compensate for incompetent drivers lack of skills, but even many of those systems raise questions. Take for example automated emergency braking and lane keeping.

    Say for example when I try to pass another vehicle, that as I approach that vehicle to initiate the pass, the system decides that I am approaching a potential collision and decides to slam on the brakes. Aside from the risk of being hit from behind by a vehicle that was going to follow me on the pass, it would encourage getting out into the opposing traffic lane too soon. I always try to stay out there for as little time as possible. Or would the system be smart enough to know that acceleration combined with a signal light means intention to pass (still a problem for some drivers who see no need to signal lane changes), and defeat the emergency braking system.

    What about when a small animal wanders out into the roadway but I’m being tailgated by an autonomous buffoon. Sorry small animal, much as I hate hitting hitting you, but I don’t want to be rear ended. The automated system would actually cause an accident. All the more a good reason to practice defensive driving techniques and encourage the tailgater to pass if at all possible.

    Lane keeping can be a challenge too, because not all roadways have equal quality lane markings. Around here, winter sanding and plowing wears off lane markings, so that they have to be repainted yearly, and that doesn’t happen right away, so for many months markings are gone or faint. Also it is not unusual to have snow cover on roadways before they are fully plowed, and often the effective lanes do not match the real lanes marked below. What happens when a lane marking is exposed and the system decides that the car is out of lane and acts to correct it, forcing the car into another vehicle’s path? An accident that wouldn’t otherwise have happened.

    Once the inevitable accidents start happening, actually caused by these safety systems, the resulting litigation will probably drive the implementation of these systems backwards.

    Improved driver warning systems may be good, but as demonstrated by my Q5’s system shutdown under certain conditions, they should only be considered as supplementary and never relied on as primary systems.

  4. jlalbrecht64
    January 1, 2018

    30 years ago at university, I created an AI program to help a user pick out electrical filters. I think of this when contemplating your point about whether to save the pedestrian or the vehicle occupants. Who’s going to decide how the car responds? Talk about, “death panels!” I see enormous potential technical problems as well as ethical and legal problems in implementation. One simple example: Would some people have the right to have a different logic in their car because they are, “special?”

    Lots of food for thought in this essay! Thanks and happy new year.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: