GPSApp update?

Hey Jan (or anyone else that might know), what is going on with gpsapp? For a while there development was speeding along, but there haven't been any new posts in a while. I was curious if you have been busy working on map data or if real life just got in the way for a bit. Anyway, I am looking forward to my first chance to use gpsapp on a longer trip for thanksgiving.

-Mike

I think Jan's on vacation for a week or so.

I have some updates overdue to him but I had a death in the family, and I'm going to London shortly.

I just used GPSApp on a long trip (8-10 hours each way, to Portland and back), and your message just reminded me to report on its usefulness.

MapsOnUs sucks.

MapsOnUs sucks big green donkey dick.

Fortunately, I knew this ahead of time and did not depend on their data. I supplanted it with printouts from MapQuest, which were excellent and provided me with the exact route data I needed.

The problem is that GPSApp is only useful to me if I can throw away the printouts and depend totally on it. Because of the fact that MapsOnUs sucks so bad, I cannot depend on its directions. Therefore, its use on this trip was merely a testing exercise.

Now, for most of the trip, which was the long haul on I-5, the map data was correct. I followed my little arrow along the line on I-5. But for the smaller highways between my house and I-5, MapsOnUs was totally wrong. MapQuest nailed it exactly, including a shortcut on a little-used road called "Woodruff Lane" which I've been using for years because it shaves at least 20 minutes off of the route. MapsOnUs had me going pretty far out of the way which would have added at least 30 minutes to the trip above and beyond the 20 minutes that the Woodruff shortcut saves me.

GPSapp itself worked quite well the whole time. However, it had a couple of problems:

- The known bug where it disappears from the Hijack menu happened a couple of times.

- The known bug where pressing the left button to zoom-out the map did not do so, instead it rewound the song with no way to stop it. This happened several times.

- It had a new bug where if you left it up and running for several hours, then for a moment the music from the player stuttered. After that, all front-panel input stopped working and I had to reboot the player by using the POWER button on the remote.

But when it worked, the basic functionality of the GPSapp software was great, it did exactly what I would want it to do. And all of the work for keeping my Rand McNally unit from going into full-reboot-mode was doing its job properly, each time I started the player the satellites came right up.

If we could clean up the aforementioned bugs, and if we could use MapQuest data instead of MapsOnUs, it would be the perfect application for me. The latter, unfortunately, is out of our control I fear.

Sorry to hear about the death in the family, Derrick.

The problem is that GPSApp is only useful to me if I can throw away the printouts and depend totally on it. Because of the fact that MapsOnUs sucks so bad, I cannot depend on its directions. Therefore, its use on this trip was merely a testing exercise.

That's the point of being able to use Census data, which shockingly seems to be more useful than MapsOnUs.

That's the point of being able to use Census data, which shockingly seems to be more useful than MapsOnUs.

Which is exactly why I was asking the original question I was wondering if Jan had made any progress with the tiger data. Anyway, thanks for the update and sorry to hear about your families loss.

-Mike

Last I knew Jan was playing with a python script to do routing based on his newfound knowledge of graph theory

I don't know where he's gotten with it.

I was just thinking that this might be very useful software for a ... I don't know ... traveling salesman?

Teehehehee. Good one, Bitt.

I jus installed gpsapp, and am replying to the only message I've found regarding a problem I'm having. GPSapp doesn't appear in the Hijack menu. I'm not seeing it as an intermittant problem though.

I loaded the Developer version of 2.00 beta13, followed by hijack, followed by preinit. I created /programs0 and loaded the gpsapp files there. I ran hack_init successfully. I created the /etc/preinit.d directory and copied M50gpsapp there. I had Hijack force the DC mode and re-booted.

I see that the player is started with -s, so the hack_init did it's job. But I don't get GPSapp in the hijack menu. Can someone point out anything I've missed?

Thanks,

Brian

Did you make M50gpsapp executeable? (chmod 755 M50gpsapp)

/Michael

Also, sometimes if you use WinZip to un-tar a file, it will automatically do a "Smart CR/LF conversion" which will ruin some types of Linux text files. I don't know if the preinit script is one of the ones that needs to be non-CRLF. But that's something else to watch out for... when you use WinZip to Untar, make SURE to turn off its "Smart CR/LF conversion" feature or you'll ruin a lot of empeg-related files when doing projects like this.

Thanks for the reply!

The preinit.d file (M50gpsapp) looks like this when I cat it on the player:
empeg:/etc/preinit.d# cat M50gpsapp
#!/bin/sh

[ -x /programs0/incar ] && /programs0/incar || exit
[ -x /programs0/gpsapp ] && /programs0/gpsapp

I gunzip'd and tar -xvf'd the file on a linux machine, so I believe the file should be OK. I also made it executable:
empeg:/etc/preinit.d# ls -la
total 3
drwxrwxr-x 2 0 0 1024 Jul 1 2001 .
drwxr-xr-x 4 0 0 1024 Jul 1 2001 ..
-rwxr-xr-x 1 0 0 110 Jul 1 2001 M50gpsapp

If you can think of something else I may have missed, I'd surely appreciate another suggestion!

thanks again,

Brian

Hey.. thanks for your reply.

I did make the M50gpsapp executable. Here's what a ls looks like on the player:
empeg:/etc/preinit.d# ls -la
total 3
drwxrwxr-x 2 0 0 1024 Jul 1 2001 .
drwxr-xr-x 4 0 0 1024 Jul 1 2001 ..
-rwxr-xr-x 1 0 0 110 Jul 1 2001 M50gpsapp

If you can think of anything else I may have missed, please let me know again!

Thanks,

Brian

Ok... trying to get a usable map.... does anyone else seem to have a problem with finding the 'raw
route description' link on the mapsonus page??

You have to turn it on under General Options, I think

I'm having the same problem, I'd set up the TTS clock last night, so I already had the preinit on my empeg... I noticed this error when starting in forced DC mode.

empeg-preinit starting...
/programs0/preinit.d doesn't exist
You might want to consider placing your startup scripts there
Using /etc/preinit.d as startup script directory

whats telling it to look at /programs0/preinit.d? /etc/preinit.d is there....

I double checked everthing... everything has the chmod 755 stuff set and is in the right place... any ideas?

I think bitt set preinit to look first at /programs0/preinit.d and /programs1/preinit.d if those are not found, it will default to /etc/preinit.d

my TTS clock still works... I'm assuming that preinit is still functioning properly... hmmm

so is that message normal?

I would say its a normal error message. I think i have the same message on my empeg, i've never bothered to format and mount the dev partition, but bitt would have the final say.

does anyone else seem to have a problem with finding the 'raw route description' link on the mapsonus page??

Derrick is correct, you must click on "General Options" then click on "Show Latitude & Longitude".

Yeah.... I finally got that.... I installed python and the windows extensions.... now it throws an error in line 4? something about modules? dunno... I have a feeling I'm doing something terribly wrong...

what's the complete error? i don't do python but i'll guess the pyhton you installed is somehow incomplete

C:\\maps>parse_mapsonus.py map.html > map
Traceback (most recent call last):
File "C:\\maps\parse_mapsonus.py", line
4, in ?
from convert import *
ImportError: No module named convert


Guess I should've stuck that on the end of the last post eh?

parse_mapsonus.py map.html > map

Have you tried it without the > map?

It automatically names it the same as the HTML file.

yeah.... exact same error

I'm not a Python guy, but to me it looks like it's looking for a module named convert - could be that it's not a part of the standard installation, or that the catalog it's in is missing from some PATH variable...

See if you can find something named convert[something] in the subtree where Python was installed... If so, then it's most likely to be a PATH problem.

/Michael

Lopan: Note that there were two ".py" files, one of which was named "convert". Did you get both of them? I have to be in the same directory with those two ".py" files when I execute the script to have it work.

Doh!!! sorry guys...

OK... so I got the whole map conversion thing.... I still can't get the gpsapp to show up in hijack... I reinstalled my developer image, reinstalled hijack.... reinstalled pre-init exactly following the www.beaglebros.com/empeg/preinit instructions... I then installed gpsapp using the instructions in the readme.... I gave M50gpsapp chmod 755, and all the other files that the instructions said needed it.... and it still doesn't work... maybe someone can make sense of this..

I see this is a developer image!
Mounting proc
Mounting first music partition
Tried to mount /dev/hda4 as reiserfs but got error 19
Mounting second music partition
Remounting first music partition read-only
Remounting second music partition read-only
Press 'q' now to go into development mode. You Have Zero Seconds To Comply...
Starting player
Starting player with -s- option


hijack: removed menu entry: "Hard Disk Detection"
player.cpp : 385:empeg-car 2.00-beta13 2002/07/24.
: command not foundpsapp:
: No such file or directoryprograms0/gpsapp
: command not foundpsapp:
Prolux 4 empeg car - 2.1434 Jul 24 2002
Vcb: 0x4086d000
! mp3_decoder.cpp :1049:Failed to find valid sync after seeking to offset 1690

Looks like you might have a DOS-formatted text file for your script.

that did the trick.... thanks Bitt

now to get the tts clock back on, and the xml interface thanks for all the help everyone

Got everything working... cool stuff... I'd also like to add.... yes www.mapsonus.com sucks!

So... I'm wondering, been reading alot about different solutions to mapping on here. I used to play around with PPC maps via streets and trips 2002. You could export small maps with pins marking your turns or areas of interest (maps exported for PPC are tiny but show restaraunts, rest stops just about everything)... and it actually had some really nice street level information... it'll let you export routes to html (however no raw data) just a pretty little maps and turns, but at 20/30 bucks it's pretty much a bargain. I'm no programmer and really have no clue as to whats needed in order to work right... but does anyone know how an app like pocket streets handles maps? and would there be any way to use that data and/or files for this? Also I remember seeing a few apps for PPC that used bitmaps or gif files (can't really remember how those worked) but I think it asked for a few nav inputs and charted you that way.

just thinking outloud....

Phew, just got back from a 2 week trip to visit my family in Holland.

Looks like a lot has been going on again. Working TTS, people are cleaning up the audio overlay patches, talk about getting IrDA up and running. Wow.

In any case, as Derrick said, I had been playing around with python scripts that pull some of the information from the tiger datafiles, and was doing some experimentation with various routing algorithms and ways to organise/index the data.

But nothing useful yet, it is still slow even on my desktop, so it will probably end up as some sort of a background thread that iteratively keeps trying to make small improvements to the 'current' route.

Oh, and real life is catching up a bit as well. I have a couple of deadlines coming up in the next 2 months that will eat up a lot of my spare time.

The known bug where it disappears from the Hijack menu happened a couple of times.

Yup, seen this as well. It happens reliably when I start the car while gpsapp is running. Hijack seems to forget that there is a third party app active, but the app is still in it's main loop waiting for intercepted buttons.

The known bug where pressing the left button to zoom-out the map did not do so, instead it rewound the song with no way to stop it. This happened several times.

Same problem. It is hard to figure out where exactly it goes wrong. Maybe the easiest way is to have hijack return an error when an application calls the WAITBUTTONS or POLLBUTTONS ioctl and there are no active redirections. That way the application knows something went wrong and can go register itself in the hijack menu again.

It had a new bug where if you left it up and running for several hours, then for a moment the music from the player stuttered. After that, all front-panel input stopped working and I had to reboot the player by using the POWER button on the remote.

Interesting, I'll see if I can reproduce this.

Interesting, I'll see if I can reproduce this.

Thanks. It was really strange when it happened because at first I thought the player had overheated and the hard disks were having problems, causing the music to stutter. Later I discovered it was somehow related to GPSapp.



In any case, as Derrick said, I had been playing around with python scripts that pull some of the information from the tiger datafiles

But doesn't the Tiger data suck, too?

I mean... MapsOnUs at least has data which includes one-way street information, and they've got a powerful server-based routing algorithm... but their directions still suck because the maps are hopelessly out of date and flat-out incorrect in some spots. So if they can't get it right, how can we expect to get it right with even-less-accurate Tiger data?

To be pedantic, it's likely to be more accurate, but less precise. Of course, that doesn't really help in practice, as the precision of one-way streets is a requirement. It'd be nice if we could figure out some way to conflate Tiger and something less precise that had one-way info.

Well, there are 2 problems:
-writing the conflation software
-finding a source of data to use once we have it

I realize that. But if there were two datasets that, when combined, provided the right amount of data, figuring that we already know how to interpret the data, we should be able to solve that problem.

However, finding a dataset that has the right missing data is probably only marginally more likely than finding a full dataset.

In reply to:

---

So if they can't get it right, how can we expect to get it right with even-less-accurate Tiger data?

---

Call it a building process. People have been talking about gps software for the empeg for at least 3 years now, but nobody has found a good data set yet. I for one am glad that has not stopped Jan so far. Here is the steps I would do if I was writing the software...

1. Build enough of a user interface shell to get data from the gps and display on screen. Done.
2. Find a source of existing routes for use on the empeg. Done via MapsOnUs.
3. Find a source of real geographic data to finish UI, this allows for info telling you where you are, showing surrounding streets and addresses, and proper moving maps. Continue to use MapsOnUs for routing info overlaid on top of the geographic data. TIGER is a good free datasource for getting this working. If this is as far as GPSApp ever gets it would still be useful for many things as previously discussed.
4. Get basic routing working with the geographic data. Get the routing algorithms and interfaces working. Find a way to get routing working with the memory and cpu limitations of the empeg (I know it can be done as I have seen Kim's software do it). TIGER data is fine for this as well.
5. Start worrying about data accuracy. Now that we have software that can actually use the data a lot more people may be willing to spend money on commercial data. Alternatively look at conflating the TIGER data (which we are now experts with due to step 3 and 4) with other freely available data. There is a wide variety of USGS data that may be usable. Also by this time TIGER2002 should be out which may be more useful to us.

Anyway, just some thoughts on why tiger is probably the best bet for now.

-Mike

In reply to:

---

Of course, that doesn't really help in practice, as the precision of one-way streets is a requirement. It'd be nice if we could figure out some way to conflate Tiger and something less precise that had one-way info.

---

IIRC tiger actually will tell us which streets are one way, but it won't tell us in which direction. I am thinking this simplifies conflation because (at least as a first step) we don't need to match everything, only known oneways.

-Mike

I've had the issue where it stutters the music then needs to get reset.... on mine also sometimes it just locks up... music continues playing but no input is accepted and the screen is frozen.

BTW heres another idea for a gps, I did this a few years ago.... if you have or can find a cheap delorme tripmate you can use these instructions for hacking it.... makes for a nice gps... since then I've modded mine to accept straight 12v power, removed the cord and replaced it with a serial port directly on the unit and a shiny LED

Also, after using this for a bit... mapsonus isn't too bad... granted I would never trust the directions its giving but last night I had a friend in the car that had no idea what an empeg was, he gets in... the voice clock starts talking... then just to demo I load the route from work to my house... it starts popping up all the CORRECT info (I think this was a first), needless to say I think he's now looking for an empeg...

OMG I didn't even notice in the link I posted above... he's got instructions for modifying his ram air intake with duct tape! I don't think this is the guy that originally came up with the delorme hack but his page was the only one I could find the instructions still intact.

Someone actually makes a frob to convert Earthmate binary to NMEA that you just place inline on the serial cable; Doing so may make an Earthmate an NMEA receiver or it may need the same hack (I don't know if it self-inits)

well... the tripmate actually sends NMEA data straight out the serial port... however it's in an off state until turned on by the software... the hack only reroutes a wire to (I think) ground a connection and change it to a permanent on state without having to use the software... I think the Earthmate and tripmate are different animals.

The Earthmate and Tripmate are different animals. The point was it's possible to modify an Earthmate to be no worse, really, than a Tripmate. The Earthmate is actually still in production. The Tripmate is not, and Delorme doesn't have any more of the clearance units either.

Ah... gotcha... how much are the earthmates?

They're over $100 new, but I'd recommend not paying that for one. If you can find one on ebay, half.com or somewhere for $50 or under, go for it.

Yeah... I think I paid around 40 or 50 for mine.... but with some of the steals and deals other people have found on the message board... definately not worth 100 bucks.

Just found this link, this guy has mapping software for ipaqlinux/yopy which uses pqa files (i think) that you can download from expedia/mapblast.... just throwin it out there

Nothing in his readmes suggest it can deal with pqa files; I should grab a fresh version and look.

The only thing that led me to that was following his instructions for downloading maps and that being the format for PDA's.... I might be completely wrong about the format.... but expedia has pretty good maps so whatever he's doing to get the info on a pda....

Such maps are not vector based but simply scrolling flat images at a fixed scale. So it really is not that interesting as you can't extract turn by turn information and other useful things.

So are most of the pocket pc apps like pocket streets and the like non vector based?

From what I've seen, yeah. Most of those are bitmaps with only the upper left and lower right corner of the bitmap being tagged with lat/long (so they can put your cursor on the bitmap from the GPS data).

My personal opinion is:
- I don't care whether the maps are bitmap or vector.
- I don't care whether I'm displaying pre-computed directions downloaded from a website, or if the directions are computed realtime on the unit.
- I don't care whether there is "live" street data that I can see on the screen for when I go off-route.
- All I care about is whether it gives me ACCURATE directions from point A to point B as I'm doing the journey.

Well... thats what i was kinda leading up to, pocket streets always gave me dead on accurate turns and such with detailed data like restops, eats, stuff like that... what we have now is great... but as long as it's usable does it matter if it's vector based or not? Many times I'd load up a streets and trips map on my PPC and it always got me where I was going... with detailed street info... it didn't tell you when to turn or anything but it was really easy to track... You could enter searches for streets and a few other cool things... other apps that work similarly do give turns... like pharos... granted this is all for windows based handhelds and such but if they are using bitmaps couldn't the same be done for empeg?

making a bitmap map useful with the limited color palette is unreaslistic.

making a bitmap map useful with the limited color palette is unreaslistic.

making a bitmap map useful with the limited color palette is unreaslistic.

Not to mention the limited vertical resolution.

With regard to mapsonus sucking... does anyone know the quality of street atlas maps? They are probably better and if you look at the general procedure this guy is following:
http://www.vsa.cape.com/~danh/gps/
You can see that Street Atlas can export a route to file and that file format looks like basically the same thing mapsonus generates with regards to lat/long for each turn. So Street Atlas costs $50 and no one even has to reverse engineer the map data to use it.. well that sounds good eh?

I'd try this myself if I had my stupid Oncore unit working, I'm not much of a solderer...