Re: Robocar--a decade away?

"Sea Wasp" <seawaspObvious@[EMAIL PROTECTED]
> wrote in message 
news:478620AE.8070201@[EMAIL PROTECTED]
> Ken from Chicago wrote:
>
>> As far robocars dealing with human drivers, who brakes faster a human
or 
>> ABS? or humans vs airbags? or humans vs a camera shutter? Human drivers

>> could be dealt with simply by raw brute machine speed and reflexes.
>
> Only if you assume the people (especially on a highway) space themselves

> out a LOT farther than normal. Because the robocar won't be able to set 
> its own spacing; if it tries to allow actual stopping distance (even for
a 
> robot), a human will cut in. Humans don't actually allow stopping
distance 
> at all; on a busy highway, there may be only a car length or two between

> vehicles, which isn't NEARLY enough even for a robot to stop in.

On busy highways at least around Chicago, there's LESS than a car length 
spacing because people will cut in front of you. Plenty of room for a
robot 
to stop in. ABS works by having the machine brake MULTIPLE TIMES per
second, 
much less simply slamming the brakes for an emergency stop.

> And if we start from the beginning (i.e., one robot car, all other cars 
> human driven) the robot car has very little chance of understanding its 
> environment nearly as well as a decent human driver.

Robocars wouldn't need to. Simple smart cruise control would prevent the 
robocar from running into anything or anyone. Airbags work because the in 
the microseconds between when a car actually does collide into an object 
sensors in the car not only INFLATE airbags to cu****on the passenger but 
also DEFLATE airbags somewhat to take some of the sting out of it.

Beat that kind of robot reflexes with humans inflating and deflating bags.



That all said, just because robocars *need* not "understand" its
environment 
to not only survive but thrive, doesn't mean it won't understand its 
environment. The problem is rather simple to break down into 3 steps:
--Conceptualize the problem: You have various vehicles and objects and 
people in an area. Vehicles tend to move backwards and forwards--altho 
having doors that can open to the sides. Some objects are static and some 
move (e.g., tall vertical objects tend to be stationary while round 
spherical objects to be mobile). People and animals are the wildcards,
most 
erratic, most irrational and can move anywhere in any direction, tho they 
have a somewhat higher tendency to follow the path of a sidewalk or walk 
*along* the street's edge. You then project "bubbles", fields around same
to 
avoid, in according to a danger / avoidance database, which is constantly 
being updated, either by downloading new data or by direct experience.
--Scan the surrounding area: Determine what in the environment comes close

to matching the preprogrammed danger / avoidance profiles.
--Process the data: Having scanned the area, you process it, analyze it
and 
issue the appropriate response to the robocar.

The last two steps is just a matter of (ever-increasing, ever-faster) 
technology. The first is a matter of, less imagination, but rather forcing

oneself to study the problem and breakdown the situation. It's easier to
do 
if you imagine a SLOW-moving robot vehicle.

-- Ken from Chicago