• Welcome to Computer Association of SIUE - Forums.
 

Robot cars talk to each other

Started by Jerry, 2005-09-20T23:01:03-05:00 (Tuesday)

Previous topic - Next topic

Jerry

Found this interesting idea in an article about the Darpa Grand Challenge:

Quote"There is great interest in car-to-car communications and crash avoidance," Rummel said.

Though technically challenging, the idea is simple, he said. What if cars could send and receive signals to disclose their presence, especially when approaching curves or corners?

Cool! So, if two cars recognized an impending crash could they take over control?
"Make a Little Bird House in Your Soul" - TMBG...

Ross Mead

The problem of car-to-car communications and crash avoidance is very interesting.  It seems a bit like the real-time systems used in air-traffic control, but adds some new interesting problems.  Let's just hope that there's a world-wide standard for the collision avoidance protocol (screw your damn "I drive on the left side of the road" or "my steering wheel is on the right"!?!).

This brings up the question of "Do you feel comfortable with the machine taking control?"  Sure, it's there to avoid a collision, but think of all that must go into making a safe decision.  Say that a communication between two cars determines that a collision may occur, and thus they plan a path to avoid each other.  But what if the path may endanger other vehicles.  Then you have a whole other problem on your hands.  There would have to be collaboration between more than just the two vehicles that currently look like they may collide.  And in that time, would you feel safer just taking control yourself?  Can a human make the decision faster?  Is the decision the "right" decision?  The computer may make the safest decision, but it may not mean completely avoiding the collision.  This could result in potential lawsuits, holding car companies or even the companies producing the algorithms responsible.  I think that this could go much deeper than at first glance...

I'm definitely looking forward to seeing the results on this one.  I'm really hoping that a team is able to get their robot further than 7.4 miles this year, but that may be a bit optimistic. :-P  My bet's on Carnagie Mellon, as they are well known for their work in robotics, and finished (well, not really "finished"... :-P ) first last year.

Geoff Schreiber

As a manager at a local auto repair shop, we were just talking about something not quite so complex today - but just simply the back-up warning systems.  If a vehicle is smart enough to beep slowly when approaching an object while in reverse, and beep even quicker as you get nearer to the object - why not just go ahead and help slow the vehicle down at the same time so that the operator doesn't hit the obstruction?  The problem is that if the computer overtakes controls - and then the human tries to, who keeps control? How long does it take to recognize the human input and switch back over to manual inputs?  What if the computer slows the vehicle down, and the MBTW (moron behind the wheel) hits the accelerator to counter-act the computer's attempts to avoid a collision?

...one day everything will be built around computer processing and the obstacle will be like a little Taz/Marvin and move out of OUR way...
~~~~~~~~~~~~~~~~~~~
Geoff Schreiber
Project Engineer
FASTechnology Group

Ross Mead

There are many interesting issues that come into play here.  Particularly, an area that is overlooked is the idea of human-robot interaction.  Of course in this case, the "robot" is the car.  How would human corrections be resolved when the machine itself makes a decision contrary to the desires of the driver?  Any corrections made by the driver would have to be able to be made quickly almost without any resistance from the machine.  And yet, the machine must be able to do it's job.  I mean, it's purpose is to take the "safest" action.  This action, however, may not be intuitive to the driver at the time.  And let's not forget the fact that humans have emotional factors that play into their decisions, while the machine may be running on strict calculations.  It would be difficult to balance that out.  Maybe a sort of preliminary driver setup would be necessary, or the car could try and "learn" via a neural network or genetic programming implementation.  Interesting stuff...

Jarod Neuner

It occurs to me that the Grand Challenge would be a pretty lowsy environment to develop this kind of technology. I mean, they are competing, correct? This standard would have to be built using the same model on all cars involved on a course.

Perhaps a better testbed would be a simulated Nascar course or maybe a one-lane loop with two cars traveling in opposite directions. That would be a bit more feasible to develop and implement a traffic control model.

Desert Traffic wha? ;p
Jarod......

Jerry

QuoteJarod wrote:
It occurs to me that the Grand Challenge would be a pretty lowsy environment to develop this kind of technology.

Hmmmm, sounds like a good project. A couple of RC cars wired with micro-controllers that can communicate with say Blue Tooth.

Anyone looking for a URA project?
"Make a Little Bird House in Your Soul" - TMBG...

Ross Mead

I'd love to do something like that...

... maybe a URA for next year... ;-)