backtop

Print 12 comment(s) - last by PandaBear.. on Nov 2 at 12:25 PM
The DARPA Urban Challenge will take place on Saturday evening

Even though the Defense Advanced Research Projects Agency (DARPA) Urban Challenge event is still several days away, teams continue to drop like flies.  Just 21 out of the 35 original teams remain in the event, after the other teams were eliminated during qualifying trials underway in Victorville, Calif.

The following teams were most recently cut:  Axion Racing, Mojovaton, Team Caltech, Team Jefferson, Team Juggernaut, Gator Nation, Team Urbanator and Golem Group.  The other teams eliminated so far can be found here.

"Vehicles are being tested in three test areas to evaluate their ability to operate with live traffic, make safe left turns across moving traffic, and pull out at T-intersections with cars arriving from both directions," according to DARPA.  "Vehicles also have to follow narrow winding roads, avoiding parked cars and other obstructions, maneuver into a designated parking spot and negotiate 4-way intersections and road-blocks."

During qualifying, the Caltech Alice vehicle cut off a human driver in an official DARPA vehicle: the first strike against her.  The next lap, Alice drove over a curb but recovered badly before ending up in a lane with opposing traffic.

It is likely a couple more teams will be cut before the race, which starts at 7:00 a.m. PST.

Six teams have already been given a reprieve until Saturday night's main event – the following teams know they will be competing on Saturday:  Tartan Racing, Stanford Racing Team, Team Cornell, Ben Franklin Racing Team, Victor Tango and CarOLO.

DARPA will hand out prizes of $2 million, $1 million and $500,000 to the first three cars that are able to complete the course in less than six hours.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
Hard programmed
By AnnihilatorX on 11/1/07, Rating: 0
RE: Hard programmed
By Samus on 11/1/2007 6:07:01 AM , Rating: 3
Depends how you look at it.

A) Humans kill other humans all the time with cars and rarely face any jail time (in the United States especially)
2) If AI kills a human, the programming manufacturer is obviously responsible
c) The programming manufacturer will point out that AI-based vehicles statistically kill far less humans than those human powered
x) They will probably be right.


RE: Hard programmed
By BladeVenom on 11/1/2007 1:33:51 PM , Rating: 2
AI-based vehicles will be sued into oblivion and we will be back to humans killing humans with cars. An AI based vehicle will always lose in court to a sobbing human.


RE: Hard programmed
By ViRGE on 11/1/2007 6:31:55 AM , Rating: 5
Keep in mind that the biggest problem for AI cars isn't really other cars, it's other humans. If driving was done by AI-only via a coordinated network, the results would be far better.

In other words, don't remove the idea of AI drivers, just remove the human drivers entirely.;)


RE: Hard programmed
By RMSe17 on 11/1/2007 11:41:42 AM , Rating: 2
This is true. The "AI" and pathfinding is too difficult to make (it's not a piece of cake, but still, look at your average video game these days, AI has few problems getting stuck on things as they did back in the early days).

The problem comes with inaccuracy of sensors and processing time. Video processing is very processor intensive, and it needs to be done in real time. Image recognition..

And then, what happens when you have your 4 sensors saying that the road is here, and one sensor saying the road is 4 feet to the right... What if that one last sensor is the most accurate? Do you trust it, or 4 less accurate ones? mmmm

too bad the team from my school got cut.. The guy who designed the circuit board for the main controller was taking the grad level AI class with me.


RE: Hard programmed
By RMSe17 on 11/1/2007 11:42:42 AM , Rating: 2
The "AI" and pathfinding is not too difficult to make.
had a typo, there should be a "not" :)


RE: Hard programmed
By djc208 on 11/1/2007 7:24:06 AM , Rating: 2
It depends, like everything else there are better applications than others. One of these won't be driving me to work any time soon but that doesn't mean that 4 or 5 years from now it couldn't be operating dump trucks at a mining or contruction site, or (like DARPA intends) delivering supplies in the desert.

Besides the problem usually isn't the decision makeing of the AI, it's getting it acurate data and processing it fast enough. There's a lot of computing power in how humans visualize their environment, even though it's almost automatic to us.

In any event it's quite a feat that this many teams are even trying, this is a very difficult problem and they should be proud they even have something workable. Just thinking about how much work would be involved makes my head hurt.


RE: Hard programmed
By Alexstarfire on 11/1/2007 9:53:40 PM , Rating: 2
It's actually not THAT much power, it's just that our brain works differently than a machines, currently at least. Our brains work by pattern recognitions and visual aids while a computer is by calculations. This is why computers whoop us in math and why we whoop computers in recognizing objects.


RE: Hard programmed
By SilthDraeth on 11/1/2007 10:10:44 AM , Rating: 2
quote:
The fact that currently we don't have the technology to make a true artificial intelligence capable of moral judgement in decision making


Good grief, I hope you do not honestly think that driving decisions should be based on moral choices?

I can see the driving test now?

1. When you arrive at a four way stop heading north, at the same time as another driver who is heading east, who has the right of way?
A) I do, because I am a good person.
B) The other driver, because I cheat on my wife.
C) I do, because I drive a Prius and the other vehicle is a Hummer.
D) How do I know if the other driver is a good person?


RE: Hard programmed
By Procurion on 11/1/2007 10:18:01 AM , Rating: 2
I am confused. You say we need '( in another words, a static hard-programmed program )' but then say we need 'really dynamic A.I. of which its decision making process is not pre-programmed rule-based '...help me out, here.


RE: Hard programmed
By SilthDraeth on 11/1/2007 11:07:18 AM , Rating: 2
I understood him as saying, we currently have "a static hard-programmed program" and that it sounds dangerous, and that we need "really dynamic A.I. of which its decision making process is not pre-programmed rule-based"


RE: Hard programmed
By PandaBear on 11/2/2007 12:25:40 PM , Rating: 2
1) In most driving accident, human is the weakest link and the cause of the accidents like distraction by cell phone, DUI, driving in unsafe speed for the condition, etc. I would trust a machine better than human in not making these mistakes.

2) The problem with AI car is not the AI, it is the SENSOR and pattern recognition of road condition. This is always the problem with computer based decision making, the hardest part is recognizing where the object is in all condition. Vision and Audio recognitions are the worst offender.

3) You think an AI driving a supply van in Iraq is not a good thing? It would cut down the number of ambushed soldiers, and reduce the enthusiasm of bombing.


"I mean, if you wanna break down someone's door, why don't you start with AT&T, for God sakes? They make your amazing phone unusable as a phone!" -- Jon Stewart on Apple and the iPhone











botimage
Copyright 2012 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki