Re: Schmidhuber

From: Brian Atkins (
Date: Fri Jun 23 2006 - 13:34:12 MDT

Michael Vassar wrote:
> What is Schmidhuber talking about when he claims that
> "In 1995, a fast vision-based robot car by Dickmanns autonomously drove
> 1000 miles in traffic at up to 120 mph. "
> Does anyone here know more? The explanation here
> gives me little understanding of why the first DARPA challenge was such
> a failure and the second such a success, nor of why there has been so
> little progress in self-driving cars in the last 11 years, nor of why
> none of the autonomous car people I knew when I worked at NIST knew
> about this.

There probably were some kind of limitations or human interactions thrown in.
Maybe they got the car on the autobahn, got it going, then turned it over to
computer control for a long straight stretch to see how fast it would go. I
haven't really looked into it though. In this article Dickmanns is quoted in the
mid-90's apparently saying that it would take more computer power to have a
truly "do it all by itself" car:

I think the first DARPA event failed mainly due to lack of testing and
development by the teams involved on the kind of real world off-road course they
had to face. Some got thrown by mechanical problems, most got thrown by their
code not being able to deal with certain kinds of challenges that the off-road
course had. With more experience they were able to refine their code to handle
all the gotchas for the second event.

Brian Atkins
Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT