Just now smart are these things?
Peter H. Coffin
of killing one and buying a model a few years old from someone else.
I've own before this a StreetPilot, a ColorMap (think SPII), and a 2620.
The 2620's autorouting was pretty straightforward. You told it where you
wanted to go, and as long as the same settings were chosen, it gave you
the same route back, every time. When there was variation, it's because
you told it "shorter" instead of "faster", or "avoid highway" or some
similar thing that you knew that you did. And if you came back to the
settings you had before, it'd give you the same route it had then
before. Sensible enough, right?
Enter the N�vi 785WT. This has all the same widgets, plus the
"eco-Route" which the manual explains as the most fuel-efficient route,
basically looking for straight lines, bigger roads, and a target speed
of about 45 MPH. Nice. Not exceptionally exciting because I drive like
an old man and the fuel difference between 45 and 53 ain't very much
anyway, but fine.
The thing that's gotten me to thinking is the 785 seems to like...
exploring. There's a particular thai grocery that I like that's about 7
miles away. With the 2620, it gave me one of two routes, depending on
whether I picked: An L-shaped route on two interstates for the fastest,
and a zig-zag route on shortest. The 785 has now made about five trips
to the grocery and is now more or less settled on a simplified version
of the zig-zag for "fastest" on the way there, and gives me a different
route for the trip home every time, still on "fastest" selection.
If I'm according the thing the maximal amount of smarts, I'd swear it
was hunting through roads it thinks are likely to be fast, then having
me drive down them to see how much time it REALLY takes to drive chunks
of them. The home routing has developed an affection for part of the
zig-zag route that it didn't start out knowing was 55 MPH (no little
signpost on the lower left), and only tried a chunk of another road
that's actually residential a single time: after stoplights ever 1/4
mile and 30 MPH speed limits, when the stretch gotten to by turning a
half-mile later is 40 MPH and stops only twice a mile, it's seemingly
decided that THAT was a bad idea, and resolved to never think about it
again. And, it's never seemed even interested in the interstate highway
option, though it's only ridden down it once.
So, does anyone know if this thing actually DOES experiment, take notes
and remember what roads seem to work well?
--
I think I'd like to see a Simpsons episode starting up with Bart Simpson
writing "I will not attempt to undermine the Usenet Cabal".
-- J. D. Falk
Happy Trails
<hel...@ninehells.com> wrote:
>So, does anyone know if this thing actually DOES experiment, take notes
>and remember what roads seem to work well?
Yes, of course it does, and it actually knows if you stopped to take a
shit along the way, and remembers where the good toilets are, hahaha!
>Enter the N�vi 785WT. This has all the same widgets, plus the
>"eco-Route" which the manual explains as the most fuel-efficient route,
>basically looking for straight lines, bigger roads, and a target speed
>of about 45 MPH. Nice. Not exceptionally exciting because I drive like
>an old man and the fuel difference between 45 and 53 ain't very much
>anyway, but fine.
If you ever have, and actually use, something that tells you your
instantaneous fuel consumption, you'll find that rate of acceleration,
coasting well before and up to a stop, and hills and head or tail
winds, make a bigger difference to mileage than the factors in that
manual.
You need to get the GPS with anemometer and pitch meter built in,
hahahahaha!
Fred McKenzie
"Peter H. Coffin" <hel...@ninehells.com> wrote:
> Enter the N�vi 785WT. This has all the same widgets, plus the
> "eco-Route" which the manual explains as the most fuel-efficient route,
> basically looking for straight lines, bigger roads, and a target speed
> of about 45 MPH.
Peter-
Doesn't the "T" indicate traffic? If the Nuvi was really smart, it
might take alternate routes when there were traffic jams.
My new 1490T did not volunteer to take alternate routes during a recent
trip across the state. Twice it warned me in advance that there were
accidents on the road ahead, but indicated the delay would be in the
order of a couple of minutes for each. Had I chosen to avoid both
accidents, the trip would have been about an hour faster!
Fred
Peter H. Coffin
> In article <slrnhgeheu....@abyss.ninehells.com>,
> "Peter H. Coffin" <hel...@ninehells.com> wrote:
>
>> Enter the N�vi 785WT. This has all the same widgets, plus the
>> "eco-Route" which the manual explains as the most fuel-efficient route,
>> basically looking for straight lines, bigger roads, and a target speed
>> of about 45 MPH.
>
> Peter-
>
> Doesn't the "T" indicate traffic? If the Nuvi was really smart, it
> might take alternate routes when there were traffic jams.
The T does indicate traffic, but only the interstate route would have
that data. And at least two trips were during times I'd not expect there
to be any traffic flow issues. It might explain why the interstates were
avoided, but not the latest variation (which prompted the question,
actually) of taking that favored 55MPH road an extra two miles before
turning south onto a designated US highway that added about ... (let's
see, sqrt of this over that...) a half mile to the total distance.
> My new 1490T did not volunteer to take alternate routes during a recent
> trip across the state. Twice it warned me in advance that there were
> accidents on the road ahead, but indicated the delay would be in the
> order of a couple of minutes for each. Had I chosen to avoid both
> accidents, the trip would have been about an hour faster!
There's certainly some lag between how fast the information goes from
whatever monitors the speed of flow, to the data publication, then to
whatever reads that and sends out the information to the recievers.
Maybe it was only a few minutes delay 10 minutes ago, before the fire
trucks arrived and blocked two lanes. (:
--
Technical points aside, you could probably beat someone to
death with a Newton if you had to. Try that with a Palm Pilot!
--Dan Duncan in comp.sys.newton.misc
who where
<hel...@ninehells.com> wrote:
(snip)
>So, does anyone know if this thing actually DOES experiment, take notes
>and remember what roads seem to work well?
Despite the cynicism of some posters, I'll relate an experience with
our 760.
Had occasion one evening to drive to a workmate's place I hadn't been
to before. It was in a "jungle" of twisty turning near-urban roads.
Part of the area I knew, a lot was new to me. Told the 760 "shortest"
and away we went.
Still in familiar territory, it wanted to send me around two sides of
a triangle. I chose the third side, which was shown on its map and
was also shorter than either of the others. On the return journey
"destination: home" it used what it had learned and sent me back the
road I had used.
Make what you will of that, and be sceptical if you wish, but it did
"learn" a shorter route.
TulsaOK
To be sure, take your original route TO your workmate's place and see
how it routes you. I've noticed that the route back home sometimes
differs from the route TO a specific place.
Mike Coon
> Still in familiar territory, it wanted to send me around two sides of
> a triangle. I chose the third side, which was shown on its map and
> was also shorter than either of the others. On the return journey
> "destination: home" it used what it had learned and sent me back the
> road I had used.
>
> Make what you will of that, and be sceptical if you wish, but it did
> "learn" a shorter route.
As a retired computer programmer, I think I'd apply Occam's razor, and guess
that it more likely to be a variety of routing "bug" (meaning in this case a
less-than-optimum calculation) which happens to show up in going from A to B
and not B to A. Even if the routing algorithms are written so that they work
from the start towards destination and from destination towards the start
simultaneously (hoping to meet in the middle at some point, of course),
symmetry would not be guaranteed.
Perhaps the problem is that the options are described as being "shortest" or
"fastest", implying perfection. If routing were described as "biased towards
..." then we would be less surprised to spot imperfections in the routes?
Mike.
--
If reply address is invalid, remove spurious "@" and substitute "plus"
where needed.
Peter H. Coffin
> who where wrote:
>> Still in familiar territory, it wanted to send me around two sides of
>> a triangle. I chose the third side, which was shown on its map and
>> was also shorter than either of the others. On the return journey
>> "destination: home" it used what it had learned and sent me back the
>> road I had used.
>>
>> Make what you will of that, and be sceptical if you wish, but it did
>> "learn" a shorter route.
>
> As a retired computer programmer, I think I'd apply Occam's razor, and guess
> that it more likely to be a variety of routing "bug" (meaning in this case a
> less-than-optimum calculation) which happens to show up in going from A to B
> and not B to A. Even if the routing algorithms are written so that they work
> from the start towards destination and from destination towards the start
> simultaneously (hoping to meet in the middle at some point, of course),
> symmetry would not be guaranteed.
Yeah, and some cheap preferences would be easy to implement and result
in different routes: for example, in North America, biasing for
righthand turns instead of left is generally going to be faster than
regarding both evenly. But that
> Perhaps the problem is that the options are described as being "shortest" or
> "fastest", implying perfection. If routing were described as "biased towards
> ..." then we would be less surprised to spot imperfections in the routes?
Which is why it'd be cleverer to make the things smarter and actually
test the gizmo's own assumptions.
(It reasserted it's liking for that 55MPH chunk of road last night...)
--
Revenge is an integral part of forgiving and forgetting.
-- The BOFH
Fred McKenzie
who where <no...@home.net> wrote:
> Still in familiar territory, it wanted to send me around two sides of
> a triangle. I chose the third side, which was shown on its map and
> was also shorter than either of the others. On the return journey
> "destination: home" it used what it had learned and sent me back the
> road I had used.
>
> Make what you will of that, and be sceptical if you wish, but it did
> "learn" a shorter route.
Who-
As TulsaOK suggested, you might take the trip again and see if the
alternate route had indeed been "learned".
Somewhere I read that the algorithms were designed to prefer right turns
and avoid left turns. Could that have made a difference in your
triangle?
I have experienced a bias against making U-Turns. In one case a
destination was on the left. It would have been necessary to go beyond
it and make a U-Turn. I was routed to the right, then to the left three
times to arrive at my destination on the right.
Fred
Howard Lester
"Fred McKenzie" wrote
> As TulsaOK suggested, you might take the trip again and see if the
> alternate route had indeed been "learned".
I did that with my Nuvi 350, and the answer for it is "no." In fact/in
addition, the Nuvi acknowledges that my alternate route shaves one minute
off the arrival time. I see that on the screen immediately after Jill lets
me know she's recalculating. I thought, after the first time I took the
alternate, that I might have had the Nuvi set for "shortest distance," but
no, it was set for "shortest time." So trying it again I got the same
result.
Thibaud Taudin Chabot
who where
wrote:
>In article <ef3hg557bop58kd67...@4ax.com>,
> who where <no...@home.net> wrote:
>
>> Still in familiar territory, it wanted to send me around two sides of
>> a triangle. I chose the third side, which was shown on its map and
>> was also shorter than either of the others. On the return journey
>> "destination: home" it used what it had learned and sent me back the
>> road I had used.
>>
>> Make what you will of that, and be sceptical if you wish, but it did
>> "learn" a shorter route.
>
>Who-
>
>As TulsaOK suggested, you might take the trip again and see if the
>alternate route had indeed been "learned".
I will certainly watch what it does as and when I visit him again.
Alternatively, I could try the return journey first and - assuming it
uses the short path - then do the outbound.
>Somewhere I read that the algorithms were designed to prefer right turns
>and avoid left turns. Could that have made a difference in your
>triangle?
Transposing left for right in the above (we are in Oz, driving on the
left ...) yes it would have - the short path was a right turn.
>I have experienced a bias against making U-Turns. In one case a
>destination was on the left. It would have been necessary to go beyond
>it and make a U-Turn. I was routed to the right, then to the left three
>times to arrive at my destination on the right.
I haven't found a U-turn bias. What I have found though is a liking
for calling for a U-turn if you overshoot a destination on a
junctionless road. And sometimes the nominated U-turn location beyond
the target is uncanny - a gateway into a large empty paddock. How the
heck do they get THAT level of detail into the database?
do...@27.usenet.us.com
> On the return journey
> "destination: home" it used what it had learned and sent me back the
> road I had used.
It just had a different idea of the proper route for two directions on what
you think of as being the same road.
Between Lower Lake and Middletown, in California, a Northbound route looks
correct, but a Southbound route will divert to a side road for a couple of
miles and come back to the highway near Hidden Valley Lake. I presume that
the side road was the highway once upon a time.
I've driven the road many times, and the GPS hasn't learned anything about
the faster route.
--
Clarence A Dold - Hidden Valley Lake, CA, USA GPS: 38.8,-122.5
Mike Coon
> ... And sometimes the nominated U-turn location beyond
> the target is uncanny - a gateway into a large empty paddock. How the
> heck do they get THAT level of detail into the database?
Alternatively someone could have built the paddock to assist all the people
who get told by their Satnavs to do a U-turn just there...
Mike Lane
> I haven't found a U-turn bias. What I have found though is a liking
> for calling for a U-turn if you overshoot a destination on a
> junctionless road. And sometimes the nominated U-turn location beyond
> the target is uncanny - a gateway into a large empty paddock. How the
> heck do they get THAT level of detail into the database?
They collect an astonishing amount of data for each road that is surveyed -
up to 214 separate attributes are collected, I understand.
I take it you've seen the Navteq surveying film?
http://tinyurl.com/3ow8cl
--
Mike Lane
UK North Yorkshire
email: mike_lane at mac dot com
Howard Lester
To follow up: apparently the third time's the charm. The Nuvi directed me
via the "new" route.