[ gpsdrive ] No developer question, but a question about an older version of gpsdrive

[Thomas Schönhütl]

Thomas Schönhütl wrote:
> Sascha Silbe wrote:
>> On Sat, Nov 03, 2007 at 04:59:46PM +0100, Thomas Schönhütl wrote:
>>
>>> /*#define ROUTEREACH (0.02+10*groundspeed/(3600*milesconv))*/
>>> #define ROUTEREACH1 (10*groundspeed/(3600*milesconv))
>>> #define ROUTEREACH2 (15*groundspeed/(3600*milesconv))
>>> #define ROUTEREACH3 (20*groundspeed/(3600*milesconv))
>>> #define ROUTEREACH4 (30*groundspeed/(3600*milesconv))
>>
>> Using the lower limits of your speed "zones", I've done a 3-point 
>> interpolation:
>>
>> #define ROUTEREACH 
>> (((1/600)*groundspeed*groundspeed+(1/20)*groundspeed+(25/3))*groundspeed/(3600*milesconv)) 
>>
>>
>> Could you test if this works equally well as your discrete approach?
>> If it doesn't work well at low speeds, could you test this 4-point 
>> interpolation (using 30 for the first point):
>>
>> #define ROUTEREACH 
>> (((1/42000)*groundspeed*groundspeed*groundspeed-(1/280)*groundspeed*groundspeed+(44/105)*groundspeed)*groundspeed/(3600*milesconv)) 
>>
>>
>> CU Sascha
>
> Hi all,
>
> it was not possible for me to go to bed without a short test of the 
> new feature, given by Sascha.
>
> Here my first impressions:
>
> - I attached the files way-test.txt and way-test.dsc which I created 
> via www.mapquest.de and my bash script
> - this route was my testing route (short distances between waypoints, 
> long distances, 30-km/h-zones, highway)
> - the first run I did with this ROUTEREACH definition: #define 
> ROUTEREACH 
> (((1/600)*groundspeed*groundspeed+(1/20)*groundspeed+(25/3))*groundspeed/(3600*milesconv)) 
>
> - I was disappointed, because the speechoutput was nearby the 
> waypoints, sometimes only 20 meters before the waypoint was reached
> - I think that the Libretto (266 MHz Pentium MMX, 64 MB RAM, running 
> GpsDrive, X-Server and festival soundserver) is at the maximum of CPU 
> and MEM usage while calculation the ROUTEREACH definition
> - I think too that some seconds move until the Libretto is able to 
> display the banner with the information and to play the speechoutput 
> via festival
> - and during all these processes the next waypoint comes nearer and 
> nearer
> - I tried a second run with this modified ROUTEREACH definition: 
> #define ROUTEREACH 
> (((1/600)*groundspeed*groundspeed+(1/20)*groundspeed+(25/3))*groundspeed/(3600*milesconv))*2 
>
> - I doubled the ROUTEREACH and recompiled gpsdrive
> - then I did the same run and now the result was nearly PERFECT!!!
> - the distance to the next waypoint was great enough to drive in a 
> secure way without problems and to have the information how to drive 
> at the next waypoint
> - only one fact could be better: at the highway with 100 km/h the 
> distance to the next waypoint could be greater (now about 300 m, 500 m 
> and up to 800 m with higher speed would be ideal and great)
> - for my useage with the Libretto the current situation is GREAT and 
> for me PERFECT!
> - someone with a more "powered" hardware should be able to verify the 
> situation if the first definition is ok with better hardware or if it 
> should be changed too.
>
> @ Sascha: Could you give me please another definition to have greater 
> distances at speeds over 100 km/h (the lower speeds 30-70/80 are ok)!?
>
> This would be GREAT!!!
>
> Thank you very much and a nice weekend,
>
> Thomas.
>
> http:/home.arcor.de/pilleus.pulcher.dsl/homepage

Hi all,

in addition to the informations given yesterday here some more experiences:

- I tested the definition of ROUTEREACH from Sascha in a current version 
of GpsDrive (svn update from today) with a 1,6 GHz Centrino notebook and 
1 GB RAM to verify the given speechoutput
- The result: I have the same behaviour as with my little and "weak" 
libretto, the speechoutput is nearby the waypoints, too late to react in 
a safe way.
- If I calculate the distance (at the moment the speechoutput shall be 
given defined by ROUTEREACH) with the calculator, everything is ok.
- Running GpsDrive the distance is not ok, I don't know why!
- The difference between calculating the distance with 100 km/h to the 
next waypoint at the moment the speechoutput comes (about 800 m) and the 
real distance while driving (200 m) is too much. This is no problem of 
the GPS unit or the processor. If I double the ROUTEREACH like I wrote 
yesterday, the behaviour of GpsDrive is better, but the great difference 
at higher speeds is the same!!!
- The option -e to use flite in GpsDrive is buggy, let me say not 
useable, because flite doesn't "say" all informations given in the 
dsc-file. I think the problem is that if flite gives an information 
about speed, direction, distance the information out of the dsc-file 
drops into the nirvana.
- It is not possible to activate the options "saydistance", 
"saydirection" and "sayspeed", because the informations are too much and 
the informations out of dsc-file are without any "hearable" result. 
There should be a "priority" setting if someone wants to use all these 
features, so that the needed informations out of a dsc-file will be 
given with highest priority.
- Festival is the better choice, even in my little Libretto. It behaves 
without problems and the speechoutput is ok, no information will be lost.

So far my experiences from today with better hardware on board ...

Thomas.

P. S.: I didn't test the 4-point interpolation from Sascha yet ...