Slashdot Mirror

← Back to Stories (view on slashdot.org)

Programming Safety Into Self-Driving Cars

Posted by timothy on from the but-just-not-plan-c dept.

aarondubrow writes Automakers have presented a vision of the future where the driver can check his or her email, chat with friends or even sleep while shuttling between home and the office. However, to AI experts, it's not clear that this vision is a realistic one. In many areas, including driving, we'll go through a long period where humans act as co-pilots or supervisors before the technology reaches full autonomy (if it ever does). In such a scenario, the car would need to communicate with drivers to alert them when they need to take over control. In cases where the driver is non-responsive, the car must be able to autonomously make the decision to safely move to the side of the road and stop. Researchers from the University of Massachusetts Amherst have developed 'fault-tolerant planning' algorithms that allow semi-autonomous machines to devise and enact a "Plan B."

10 of 124 comments (clear)

  1. they're a disaster by slashmydots · · Score: 4, Insightful

    If you're not aware of the level of performance of current self-driving cars, let me break it down for you. They can't stop for construction or understand rerouting from it or obey temporary signs. They can't see stoplight colors while the sun is setting anywhere near behind them. They can't drive on snow at all. They will slam on the brakes for a piece of newspaper blowing across the road or other low density objects. They think puddles are obstructions and will slam on the brakes.

    They're basically deathtraps on wheels and they don't work at all plus they're illegal in several states.

    1. Re:they're a disaster by turkeydance · · Score: 4, Insightful

      that resembles quite a few drivers.

    2. Re:they're a disaster by ShanghaiBill · · Score: 5, Informative

      They're basically deathtraps on wheels and they don't work at all

      SDCs have already logged hundreds of thousands of miles on public roads, and have a safety record better than human drivers.

    3. Re:they're a disaster by burtosis · · Score: 4, Insightful

      They're basically deathtraps on wheels and they don't work at all

      SDCs have already logged hundreds of thousands of miles on public roads, and have a safety record better than human drivers.

      Highly misleading comment. Those tests have been on perfect condition roads, pre-planned everything, no construction, no rogue animals or children, no snow, no lose dirt or gravel, hell i doubt it was during bar close. Compare apples to apples please. Compare straight driving on highways and roadways under perfect conditions to humans and i doubt AI is better. Compare AI to humans in adverse conditions and it's like comparing a drunken teenager getting road head while texting to, well damn near anyone sane.

    4. Re:they're a disaster by Anonymous Coward · · Score: 3, Insightful

      Those tests have been on perfect condition roads, pre-planned everything, no construction, no rogue animals or children, no snow, no lose dirt or gravel, hell i doubt it was during bar close. Compare apples to apples please.

      He did compare apples to apples - the SDC outperforms humans driving in the same conditions in which it was tested. The fact that they haven't yet been tested in other conditions doesn't reduce the significance of that fact.

      Compare AI to humans in adverse conditions and it's like comparing a drunken teenager getting road head while texting to, well damn near anyone sane.

      Pure speculation.

    5. Re:they're a disaster by burtosis · · Score: 4, Informative

      I worked for 7 years in a robotics lab so i do know a few things about vision and vehicle automation. What grinds my gears about this is every last mile was pre planned. Routes were mapped in gps, every last sign, stoplight and speed limit was pre-programmed in. Every single test was on a sunny day with free flowing traffic. Even under those circumstances the algorithms spazzed out and did very unhuman like things. Sure it sounds nice to lock up the brakes for a blowing trash bag but that's asking to be rear ended and is highly dangerous.

      TL:DR they took ideal conditions under which normal humans fare far far better than on average and ran their AI. They then compared this mean time between failure to what humans have to deal with on average in totally different enviornments - rain and snow - asshole drivers in traffic jams, unexpected icy conditions - drunken driving. It's not science it's intellectually dishonest.

  2. Problem. by khasim · · Score: 4, Insightful

    "People are unpredictable. What happens if the person is not doing what they're asked or expected to do, and the car is moving at sixty miles per hour?" Zilberstein asked.

    So the car is travelling at 60 MPH on automatic when a situation arises that requires the car to switch to human-control ... and there might be a problem with the human not reacting correctly?

    I think that the problem would be expecting the human to take control and do anything useful at that speed if the programming couldn't handle it.

    1. Re:Problem. by eth1 · · Score: 3, Insightful

      "People are unpredictable. What happens if the person is not doing what they're asked or expected to do, and the car is moving at sixty miles per hour?" Zilberstein asked.

      So the car is travelling at 60 MPH on automatic when a situation arises that requires the car to switch to human-control ... and there might be a problem with the human not reacting correctly?

      I think that the problem would be expecting the human to take control and do anything useful at that speed if the programming couldn't handle it.

      It more like it's unreasonable to expect a person to be able to sit and pay enough attention to what's going on when they're not engaged in the task at all. I either want full control, or no responsibility for control.

  3. Re:AI is too unreliable by vakuona · · Score: 3, Insightful

    They make the mistake of thinking that you can get to self driving cars with a lot of miniscule improvements on current technology such as automatic braking and cruise control. A self driving car is an entirely new paradigm, much like the horseless carriage was a completely different paradigm. If you want to make a self driving car, then the working assumption should be that it has one mode - self driving. Actually, imagine the car without a steering wheel, no accelerator pedals or brakes. Imagine the car going round town with no driver in it. If the failure mode of your imagined self-driving car requires a driver to take over, then you have failed to create a viable self driving car.

  4. Re:AI is too unreliable by vakuona · · Score: 3, Insightful

    It's hard enough for human to keep attentive on the road when they are fully in control of the car. Can you imagine humans having to take over when something has failed. By the time the human being realises that their car has failed and they are required to take over, they will have crashed already.

    "Human taking over" is a really really bad failure mode in a self driving car. It's way worse than the computer trying to take appropriate action to prevent accidents and loss of life.