"karl gruber" wrote in message
...
Why does the 396 in "Automotive" mode sometimes take me off an interstate
off ramp, and then right back on the on ramp?

Best,
Karl
ATP, CFI, ETC.
World's most hangar queeny Skywagon
http://temp.corvetteforum.net/c5/kgr...fterwaxjob.jpg


"Travis Marlatte" wrote in message
t...
I work for NAVTEQ - although I am not directly involved in the data



I can share some of the issues that I am familiar with.

Navigation systems are not running on the most powerful computing platform.
So, anything that they do is a tradeoff between response time and quality.
To find the truly optimal path between two points on the map is
computationally prohibitive.

The Route Calculation algorithm wants to find the least cost path from point
A to point B. Lease cost may be defined as quickest based on predicted
travel speed or smallest distance - combined with other possiblities like
avoid tollways, avoid highways, etc. Every road in the database has
attributes, a length and a speed. Even the intersections have tables to
indicate the "cost" to get from one branch to another through that
intersection.

Techniques are used to reduce the amount of searching that must be done but
these techniques sometimes result in far from optimal results. One of the
popular algorithms that is used searches from both the origin and from the
destination. During the search, branches are pruned off that don't seem to
be going in the right direction. This, of course, sometimes eliminates one
of the good routes. The other result is that sometimes the two searches meet
but because of the ordering of the search paths, the first meet is not the
best but somehow wins.

Of course, the other problem is personal preference. My sound-bite is "only
use a navigation system when you don't know where you are going and you will
be please with the result. Use it to get to the office and home every day
and you will soon conclude that it couldn't route itself out of a paper
bag." It's going to be a long time before navigation systems will cut down
that side street to avoid that typically long light.

To address your specific question of ramp jumping, fundamentally, the
algorithm believes that getting off and back on is shorter (or faster) than
staying on the highway. I have seen several causes: 1) the user has chosen
to avoid highways but there is no other way to get there. But, while
determining the path, the algorithm mistakenly is still trying to avoid
highways wherever it can; 2) for some reason, the cost of staying on the
highway is less than using the ramp. This certainly could be a mistake in
the data but it could also be a bug in the algorithm. 3) In pruning, the
algorithm eliminated the path that stayed on the highway but then was forced
to make a connection and the ramps were used instead.

Another problem is that ramps are typically posted at something like 45mph.
I know that, for a while, the algorithm we used presumed that people would
drive them faster than coded which basically made them as good as a highway.
Prune a few branches. Ignore some data. And, presto, the ramp is better than
the highway.

If you happen to remember the highway junction where that has occured, I can
check our data.
-------------------------------
Travis
Lake N3094P
PWK