Okay, so I've got some friends who are moving into a new house next week and I wanted to see what it looked like on the way to work this morning. So I punched the address into MapsOnUs and converted it into directions for GPSapp.

GPSapp worked flawlessly and got me there.

However, the MapsOnUs directions sucked completely. And they sucked in a very unusual way. The VECTOR ROUTE itself was perfect. By using GPSapp to follow the lines on the screen, I got there perfectly.

But the TEXT DIRECTIONS that accompanied the vector route were completely wrong. They were constantly telling me to execute a "turn" at a waypoint onto a new street when all it was was a bend in the road (with no intersection). The street names were completely wrong almost the entire route. And in certain places they were telling me to turn RIGHT when the turn was clearly (and visibly, on the vector data) a LEFT.

I don't see how their supplied vector data for the roads could have been so perfect and accurate, yet the text directions could be so wrong.

I don't see how their supplied vector data for the roads could have been so perfect and accurate, yet the text directions could be so wrong.

I do. I find a Euro-centric map site a few days ago that included maps of the U.S.

It had an error I saw on mapsonus a few weeks ago (but I guess someone since complained about)

It picked the wrong fork on the PA turnpike at breezewood and had you get off, then get back on at the next exit.

I suspect they're using raw or perhaps out of date TeleAtlas data.

Ok some of this might be my fault, some of it really is the quality of the maponus data. If you look at the route file that is put on the player you'll notice that I actually strip most of the 'turn left/bear right/continue' text. And regenerate them at the time a turn comes up by looking at the incoming and outgoing heading.

You should probably check the original directions. I tried to get things to match up as closely as possible (they are probably using the same trick as I am), but I might have gotten a sign wrong, or a rounding error.

I can't autogenerate streetnames, so anything that is wrong that area is their fault

can gpsapp be modified to show a base map so you actually know where you are? i can imagine it scrolling like the Image Pan visual

I assumed the text LEFT and RIGHT directions were straight from them, so okay, that could possibly be part of the problem. You sometimes tell me to turn right when it's clearly a left at a 90-degree T intersection (not ambiguous).

Do you choose the waypoints yourself? Because this thing had waypoints that shouldn't have been waypoints. Places where I was staying on the same road (and the vector data was correct showing me staying on the same road), but it was telling me to make a turn.

can gpsapp be modified to show a base map so you actually know where you are? i can imagine it scrolling like the Image Pan visual

I'm sure it could, but that would confuse me on the low-rez screen. I really liked having just my target track on the screen, made following turns and such very easy.

I'm looking at Tiger/Line datasets to get street level information. But the small screen does get overcrowded really quickly, and the datasets are huge, so I'm looking at ways to speed up finding all line segments in a given area (spatial indexes).

Eeew. Tiger. Yuck.

I guess it would probably go wrong easily when we're close to the waypoint due to positional errors. Maybe I shouldn't look at where the current position stands compared to the waypoint, but look at where the current 'vector' would cross the line segment that comes right after the turn.

The waypoints themselves are not chosen, they are the same ones that mapsonus gives. It seems like they pretty much give a new description whenever the street description changes. i.e. I've seen "Follow Penn Ave (Rte 8)" "Continue on Penn Ave". Still the same street, the description just changed slightly.

Most of the maps that you would have to pay a lot of money for are nothing more than cleaned up tiger/line maps. They just added one-way street info and what intersections are a bridge or tunnel in real life.

Still the same street, the description just changed slightly.

For this morning's route, I was on the same road the whole way, and the vector data was correct for every twist and turn of that road. However it kept chaning the road name completely. Not a slight description change, I mean the whole road name changed. For instance, it changed from WALKER DR to UNNAMED ROAD to RIDGE ROAD and back again several times. When the entire time, it was all the same road and it wasn't even any of those names (it was actually "Rough and Ready Highway").

I'm sure it could, but that would confuse me on the low-rez screen. I really liked having just my target track on the screen, made following turns and such very easy.

Turning stuff like that off is what options are for. Turning it on is what development is for.

Eeew. Tiger. Yuck.

It's that, USGS DLGs, or pay for data that no format is available for.

Sounds like a no-brainer to me.

In reply to:

---

Most of the maps that you would have to pay a lot of money for are nothing more than cleaned up tiger/line maps. They just added one-way street info and what intersections are a bridge or tunnel in real life.

---

Yeah, but one ways become a lot more critical than you may think. We went through this with Kim Salo's mapping software. If one ways only affected things like driving in SF where there is a lot of one way streets, but the next street is most likely a one way in the other direction it wouldn't have been that big a deal. Where it became a real problem was freeway on/off ramps. Those are all oneways and the tiger data (apparently) doesn't know which way they are going or which side of the freeway it was attached to. The end result was that more often than not it would tell you to drive passed the offramp you wanted and then tell you to turn across traffic (on the freeway!) to exit via the onramp on the other side of the freeway. Not very helpful. IIRC that was another free dataset provided by one of the US bureaus (SDL, SDAL, something like that) with accurate one way info, but no street names that Kim was working with at one point. If somebody could manage to compine the two datasets accurately that would get you the one way info.

-Mike

free dataset provided by one of the US bureaus (SDL, SDAL,
something like that) with accurate one way info

USGS. The datasets are DLG. I mentioned them elsewhere in this very thread. In Kim's thread we talked about conflating these same 2 data sources, also.

Thank You Jan for developing and releasing this truly awesome application for the Empeg. Now I can use my extremely cheap GPS (Garmin 12) for more than just finding my way out off the woods when I'm riding my dirt bike. You and the others like you that are so commited to the Empeg really do bring an added value to it. Thank You and the others that helped develop this application.

For instance, it changed from WALKER DR to UNNAMED ROAD to RIDGE ROAD and back again several times. When the entire time, it was all the same road and it wasn't even any of those names

You sure you wern't teleported to Colorado Springs for a bit? There are many roads here that randomly rename themselves. For example, one I know of goes like this:

Fontmore, Fontanero, Fillmore, Circle

Then there is Delta, that manages to cross a major road three times for some reason.

You sure you wern't teleported to Colorado Springs for a bit? There are many roads here that randomly rename themselves.

Heh, no. Around here, there are sometimes roads which do that (multi-segment roads that change names along their length), but I wasn't driving on one of those roads.

And it's not just a question of outdated map data, because the road in question in that particular example hadn't even been a multi-segment road ever, at any time in the past. It was always the same road with a single name, for as long as it's existed.

I mean, I could even understand if the MapsOnUs software got briefly confused and put me onto a side road or something. But as I watched the track on the screen, it literally followed every twist and turn of the road perfectly, my position indicator was very accurately tracking on the road that was programmed in with the lat/long data. So the data for the actual road itself was correct. It's just that MapsOnUs seemed to think it was a bunch of different roads, with waypont turns several times along its length.