05 February 2018

Will Robots Replace Riders?

Maybe I hang out with all the wrong people:  I have yet to meet anybody who likes the idea of a driverless car.  They may not enjoy driving, but they are skeptical that a computer program, or whatever would guide the vehicle, can make the same kinds of judgments a driver could make.

Then there are those people who enjoy driving.  I don't imagine many of them would be crazy about losing one of their pleasures.

So why, then, would anyone want to teach a computer how to ride a bicycle?

Computer scientist Matthew Cook, from what I can see, isn't trying to make a machine that can usurp the role of a cyclist.  Rather, he says, "we do not have great insight into how we ride a bicycle" no matter how well we may ride.

 

In 2004, when he was at the California Institute of Technology, he created a simulator and made 800 unsteered runs with it to see how far it could go when there is no one to steer it.   The image above shows the tracks of those runs, initiated when the "bicycle" was pushed to left to right, and how far they went before falling down.  Oscillations from side to side, visible in the chart, occurred because the bicycle was moving too slowly to keep itself stable.

As a result of this work, he found that it took a simple network of only two neurons to keep the bike stable:  one to calculate the required lean of the bicycle to execute a given judgment in direction, and another to translate that change into an amount of torque to apply to the handlebars.

Cook says his work could have "many applications", but doesn't specify what they are.  My guess is that it might be helpful for people to regain skills and faculties lost or impaired in crashes and other traumatic events:  Simulators like Cook's might, for example, provide insights into how our minds and bodies allow us to do some of the things--like balancing a bicycle or walking--we do instinctively. Also, I could see how "test dummies" for bikes could be developed to better test helmets and other products.

I just hope no one develops robots that can push us aside and take our bikes!  

4 comments:

  1. I used to enjoy driving, then . again this used to be a fairly empty part of the world. Most drivers skills, better to say person sitting in front of the steering wheel, are stupid, careless and inept. There is a good chance that driverless cars could eventually be a good thing for cyclists. No more lunatics swerving towards us "for fun", the dumb ones who say"I didn't see the cyclist" will be head down twiting or playing stupid games on their mobile toys well out of the way.May not be this year but the cars will get better very quickly now.

    ReplyDelete
  2. Coline--You make a good point: Perhaps "driverless" cars will eliminate the stupidity,er, human error in driving!

    ReplyDelete
  3. Unless they are made illegal (after a ^&*% storm of protest), cars that predate the electronic era will continue to use our roads. I don't think my 1967 Jaguar can be made driverless. It doesn't even have power steering or electric windows! One good thing about driverless cars is that the producers of same have added liability if their product doesn't work compared to the one who turned the thing on that may not have insurance, assets or even a license. Still, cyclists will not be able to assume that old Chevy truck is officially driverless...

    ReplyDelete
  4. Steve--I'm not sure that I'd want to make your 1967 Jaguar driverless!

    You make an interesting point about liability. I wonder, though, whether those companies would be better able to protect themselves against liability simply because they have more resources than individual drivers--and, if they become big enough, more than the insurance companies.

    ReplyDelete