Jack May
2008-08-05 22:23:44 UTC
The HDTV 4th hour show of the "DARPA Grand Challenge" of testing robotic
cars with no drivers was last night on the Science Channel. Carnegie Melon
won 1st for $2M, Stanford second for $1M, and Virginia Tech in 3rd for half
a $M. True to California culture, the Stanford car was the fastest to cross
the finish line.
Six of the 11 finalist finished. The race was in Victorville Ca which is
an abandoned military base with a full community of buildings and roads. So
we had ghost cars with no drivers in a ghost town with no residents. Sort
of spooky :-)
I was surprised how well the cars worked in a complex environment. They
were all programmed to follow the rules in the California drivers manual
that we have to study to get our driver's license.
There were cars driven by people, the judges, The robot cars were able to
drive well even in a mixed robot and people traffic situation. The robots
even inherently caused traffic jams at times. There was even a dirt road
which caused problems for some cars.
There were also "police" that gave out tickets to the cars. Those tickets
counted in reducing scores used to determine the winner. Around 2012 there
will be production cars with digital car to car communication which will
drivers warnings to prevent hitting other cars.
If there is an accident, it will be interesting to see how the police handle
tickets. Could our cars get tickets instead of us at times? A cars
become automated over maybe the next couple of decades and do the driving,
could they accumulate tickets for bad driving?
The first accident between two robot cars occurred and was caught on video
for replaying many times in the future as part of stock clips. One robot
car stopped trying to figure out what to do next. The Stanford car came up
behind it and decided the car had quit running. The Stanford car then
decided to go around it. Just as the Stanford car got almost around and
turned to get back in the proper lane, the other car made up it mind what to
do and started moving. The car ran into the front wheel region of the
Stanford car.
This was unpredicted emergent behavior. DARPA did not disqualify the cars
because this challenge was not testing for this type problem. Maybe in a
future test.
cars with no drivers was last night on the Science Channel. Carnegie Melon
won 1st for $2M, Stanford second for $1M, and Virginia Tech in 3rd for half
a $M. True to California culture, the Stanford car was the fastest to cross
the finish line.
Six of the 11 finalist finished. The race was in Victorville Ca which is
an abandoned military base with a full community of buildings and roads. So
we had ghost cars with no drivers in a ghost town with no residents. Sort
of spooky :-)
I was surprised how well the cars worked in a complex environment. They
were all programmed to follow the rules in the California drivers manual
that we have to study to get our driver's license.
There were cars driven by people, the judges, The robot cars were able to
drive well even in a mixed robot and people traffic situation. The robots
even inherently caused traffic jams at times. There was even a dirt road
which caused problems for some cars.
There were also "police" that gave out tickets to the cars. Those tickets
counted in reducing scores used to determine the winner. Around 2012 there
will be production cars with digital car to car communication which will
drivers warnings to prevent hitting other cars.
If there is an accident, it will be interesting to see how the police handle
tickets. Could our cars get tickets instead of us at times? A cars
become automated over maybe the next couple of decades and do the driving,
could they accumulate tickets for bad driving?
The first accident between two robot cars occurred and was caught on video
for replaying many times in the future as part of stock clips. One robot
car stopped trying to figure out what to do next. The Stanford car came up
behind it and decided the car had quit running. The Stanford car then
decided to go around it. Just as the Stanford car got almost around and
turned to get back in the proper lane, the other car made up it mind what to
do and started moving. The car ran into the front wheel region of the
Stanford car.
This was unpredicted emergent behavior. DARPA did not disqualify the cars
because this challenge was not testing for this type problem. Maybe in a
future test.