Thread: BETA testers needed for StreetDeck Europe

The MapPoint map is more accurate - the route lines correspond to roads on the Mappoint map, it's just that the roads on the Virtual Earth map don't map onto the roads from the Mappoint map.

Using GPS in Mappoint gives exactly the same location as StreetDeck (with the GPS dangling out of my office window )

Below are the Virtual Earth and the Mappoint version of exactly the same location. If you superimpose the 2 images, you can see that they don't quite match up.

I'm using Mappoint 2004 btw.

... finally bright spot in the skies, so I'm back on the road.

No GPS fix loss ocurred this time and I could perform short-term (even within 15 min life span) driving test.

GPS road matching exposed good results (usually 1-2m). Some mismatches (2-5m) I'd rather put on MP map digitizing errors.

As already noted by Teranga, there is small error (like a delay in displaying current position) that is dependent on current speed - approx. 20-30m per each 50 mph. What's interesting, very similar error presents in my in-car nav, too (I've been always thinking that these corrections are not too hard to implement).
While insignificant on slow speeds, it migth be confusing, for instance, while driving highway and having map zoomed to certain extent - exit is not shown on the map yet, but car actually already has passed it...

regards,
Tomcat

When you open the serial port of the gps device with hyperterminal for example, it shows updates about every second and it seems that the current position is not always sent within that second. Then between 1 second or even >=2 seconds, there is time for position error. We could compute distance error with following formula:

d_error=speed/3.6*refresh_rate

where:

• d_error is distance error
• speed is given in km/h
• refresh_rate is time between each position update in second

for example with the given information from tomcat:

Originally Posted by tomcat

approx. 20-30m per each 50 mph

tomcat: did you really mean mph or km/h?

We can verify the formula with a refresh rate from 1: d_error=50/3.6*1~=14m
or if 50 was really given from tomcat in mph: d_error=80.5/3.6*1~=22m
if we double the refresh rate, we double the error. Results seem to correspond to what tomcat wrote.

From my earlier posted information, I said that with a speed of 240km/h I was estimating the error of about 300m-400m, but thinking about it again, I would re-estimate it down to 100m-200m.
d_error=240/3.6*1~=67m and I think that the position refresh rate of my gps device could be within 2 seconds, then the error could be about 140m, which would correlate my new estimation.

Then supposing that my therory is right (some gps guru could tell me if I did a mistake...), how could a navigation system shows correct current location when the current position does not come fast enough from gps device? Then neither SD nor MapPoint could do something. Apart of estimating the position depending on d_error when the navigation is based on a calculated route...

Why is refresh rate so slow? is it because of satellite signal? is it because of gps device? For satellite signal, we could wait for Galileo, maybe it will be better than gps in that case, or in the case of gps device, why not building faster devices???

Originally Posted by Teranga

tomcat: did you really mean mph or km/h?

Well, at second thought I can correct it to ~40 mph or ~70 km/h...
But I was just referring to the directly proportional nature of the delay from speed.

I guess the actual delay coefficient is not universally constant but rather depends on round trip time of the received GPS position until it is displayed on the map (including computing, map update times, etc).

Had a thought about the Virtual Earth misplaced images - could it be because satellite images tend to be lower resolution that US images? Just wondering...

When swithcing Navigation view into Browse mode, touching and/or clicking the point on the map suppose to move that point into center of the view, imho.

The actual results are:

• in StreetDeck
  
  1. Select actual point
  2. If there are some POIs in certain proximity, select one of them (probably first in some list)
  3. Jump somewhere (could be kilometers away) that at first sight looks like random location (actually, jumps to location from the list described in the p 3. below)
• in MapPoint
  
  1. Select actual point
  2. Select POI if there is one in certain proximity (display a list if there are more then one POI)
  3. On single mouse click does nothing, on double-click presents list of geographical- and administrative units the point belongs to.
     Note: the point SD jumps in p. 3 above is the first administrative unit from this list

I hope that explanations are enough, 'cause making illustrations is a bit annoying.
Tell me if you need one

Chuck: tell me if you need more info and logs about this

Originally Posted by tomcat

...I guess the actual delay coefficient is not universally constant but rather depends on round trip time of the received GPS position until it is displayed on the map (including computing, map update times, etc).

That's correct therefore I wrote the constant "refresh_rate". This can be a pseudo constant depending on how frequent the navigation system is receiving positioning data from gps device.

Originally Posted by god_of_cpu

There is a bit of a delay with GPS smoothing on. If you turn this off in navigation settings it make the GPS more accurate, but choppier.

Having switched the smoothing off, the distance error is smaller, I would say about half of it was.

Originally Posted by god_of_cpu

StreetDeck doesn't yet support Unicode characters, but will at some future date.

On the system I have currently installed the sd_europe.zip (2nd version), I do not see "ü" in "State" field... but I remember that I had the "ü" on another system which could be more "germanized" as the first one. Could it be true or I dreamed? I will check this again. Are european accents not part of "normal" ASCII table and do not need UNICODE program to display them correctly?
ms-help://MS.MSDNQTR.v80.en/MS.MSDN.v80/MS.VisualStudio.v80.en/dv_vsref/html/55a5ff74-169c-49d9-9d5d-5e6cba634ac5.htm
(MSDN keywork: ASCII character set)

... a little test later: I can confirm that on the other system, the State field is automatically filled with "Baden-Württemberg" with "ü" instead of "u"... so how can it be? the other system has same regional settings, same winxppro sp2, same mappoint europe 2004!

Originally Posted by Me

Next test steps:

1. Driving thru german-switzerland border to see whether it won't be confused (will be a little bit difficult getting thru the border in 15 minutes... but I will see...)
2. Doing bigger rerouting (if possible).
3. Adding waypoints.
4. While driving, finding nearest hotel.
5. Any requests from Chuck of course

Results:

1. Passed. No problem found. I remember that some navigation system have problem with it.
2. The calculated route told me to turn left and I turned right (on an highway to another one), then SD recalculated the route from my current location till next highway exit to come back to the correct highway to correct my simulated driving error. The additional trip was about 30km. SD did everything correctly.
3. I didn't do it... next time
4. Didn't get any I presume Chuck is overloaded with our log files or with some other important tasks or he has already all informations he needed to know Do not hesitate!

Remarks:

1. I was searching before I did my trip from Germany, my swiss address, I only entered "Biel" and SD was always stuck to results from "Germany" this time.
2. Are highways black? Mappoint does it differently (red fat line with yellow line in the middle)... I remarked that some areas are also completely black whereas it is grey with mappoint (lat. 47.532, long. 7.661).
3. Every weekend I drive from germany to switzerland and back... this is about 350km one way. Today it was very hard to test with 15min limitation and I think it would be interesting for you Chuck to test a longer european trip for stability (crash, memory leaks, ...) and reliability.

Originally Posted by Flibblebot

...
Below are the Virtual Earth and the Mappoint version of exactly the same location. If you superimpose the 2 images, you can see that they don't quite match up.

I'm using Mappoint 2004 btw.

Flibblebot, are those two images, you've mentioned, from "exactly the same view"? I mean, is SD's Zoom factor the same?
I'm asking this because the scale factor of two maps (MapPoint and Virtual Earth) obviously is not the same - try to compare lenght of different segments of the roads...

This, probably, should be addressed by Chuck - to make sure both maps are scaled to the same extent.

I'm trying now to compare the Virtual Earch, Google Earth and MapPoint representation of the same location.

Immediately a question arose: is there a way to enter latitude-longitude coords into Virtual Earth or, at least, have coordinates of the point under cursor (or other) displayed?