Road Types (USA) – comprehensive overhaul of drivable roads

[bgodette]

Yep, I've looked at the FC map. As I've said I see the idea behind the FC maps, but don't really see the need to change thousands of miles of roads across multiple states. IF and only if, Minor does still get the same long range routing benefit that it used to. If that has in fact changed then I stand corrected and will start tasking my AM's to start reclassifying (Yay edit counts climb!)

The only thing that's changed with how routing has worked since ~2009 is Private/PLot/ToD.

[bgodette]

Something has changed. As I said, my previous test route is no longer working.

Something may have changed, but it wasn't with how routing has been working for the last 4+ years.

[bgodette]

But my point is made: 99% of the time, it's going to stick to the major highways like glue, and only if there is a SEVERE (subjective term) amount of traffic on the major highways will it consider other options.

Basically, if a non-major highway is the faster route with ZERO traffic on both that route or the major highway, it will STILL choose the major highway....because it's a major highway. It's only if the major highway is actually experiencing traffic that it doesn't use the major highway.

Except that's not the experience here, for routes that are quicker using non-MH/Freeway it will happily give them to you. Those instances are few and far between simply because anything that's used for serious connectivity is generally already NFC'd as an arterial or expy or interstate.
Castle Rock outlets to Parker
Fairplay to Georgetown. This one is interesting because option 3 is a mountain pass that's typed as Street. Approximately 23.6 miles of Street once it departs US-285.

Experience says otherwise. In fact, so does the wiki! The entire reason we set at-grade connectors to have the same type as the roads they're connecting is because of penalties in changing type. Same with the instructions on what type to use for roundabouts depending on what roads they connect. If there wasn't a routing penalty for changing types, we wouldn't have to do that, would we?

The wiki is wrong if it says that. There are no penalties for changing type except for Private/PLot to anything else. Most of the type rules for connectors and roundabouts are mostly about aesthetics and not anything to do with how routing works, and most of them were written before anyone knew anything about the actual data model being used. Most of what we know now is only known because some editors developed ways to actually test the assumptions we thought were true or believed because someone heard from someone else who heard from yet another person who interpreted what was said in passing in an email and it's gone through 4-5 language translations.

[bgodette]

Why do so many of the discussions on here turn into "The wiki says this" with a response of "The wiki is wrong"?

The only way I have to prove anything is anecdotal experience and wiki. If the wiki is wrong, then fix it...that's why it's a wiki, so people can fix it and keep it up to date. Why I get 'the wiki is wrong' so much, I don't know. >.<

In any event, here's at grade connectors: https://wiki.waze.com/wiki/At_grade_connector#Road_type

And roundabouts, which goes into much detail about the routing penalty: https://wiki.waze.com/wiki/Roundabout#Road_type

As I said "if the Wiki says that it's wrong", so as expected, the Wiki doesn't say what you seem to think it says.

At grade connectors mentions nothing about type transition penalties, and Roundabouts links to a page that describes the penalties that do exist, which does NOT include any general type transition penalty which only exists for Private and PLot.

The rules for at-grade are aesthetics, it generally looks better when done that way. Same thing for roundabouts. The UK likes the aesthetics of highest type wins for roundabouts over the NA rules. Both rulesets simply set a consistent look, both route equally well simply because long distance pruning never gets aggressive enough to exclude such a limited run.

You're getting hung up on pruning vs preference. There is no such thing as preference or stickyness in Waze's routing. There is pruning, but that's a fact of live in any routing system, even Google Nav's.

The differences you may see for long distance routing between Google Nav and Waze invariably come down to Waze's map not being consistent with NFC while Google Map/Nav is. Every single "Waze doesn't give me XYZ route" problem I've ever worked on has come down to a problem of Type and connectivity. Type to NFC and amazingly you get the exact same routes as GMap/Nav outside of general traffic avoidance (which Waze does better).

[bgodette]

A lot of the text in the wiki is old and has survived revisions without comment. It's really pruning we're trying to avoid, in case a roundabout is in the middle of a route. Same as at-grade connectors.

Not to mention that this silly argument is about typing of at-grade connectors and roundabouts is minutia that is irrelevant because the number of segments and distance involved isn't going to exclude it via pruning.

[bgodette]

So much fun discussion! Did I miss any routing test users chiming in that the new routing code does a much better job of using mH and primary street, using speed data rather than pruning based only on road type?

The differences have been minor for me. It mostly comes into play for "last mile" routing on my commute since I'm 80% freeway to begin with unless it's a bad snow day. In adverse traffic conditions it has given me a route very much like this. With the exception of a short bit of I-225 N, that's all Primary, but heavily trafficked. I'm not sure if the non-rates routing would have given that or not.

[bgodette]

I tried all the examples from [url=https://www.waze.com/livemap/?zoom=15&lat=45.78328&lon=-108.50068&from_lat=44.08054&from_lon=-103.23101&to_lat=45.78329&to_lon=-108.50069]this post plus the route from Rapid City, SD to Billings, MT with the routing test, but all the results were the same in every case.

Yeah, that's disappointing. Did you include this example in the routing test feedback thread?

There's a missed restricted turn preventing it from working, posted in the other thread.

[bgodette]

To stir the pot a little bit, have they actually came out and said that the reference speed is now the speed used to calculate the road and road type is no longer factored in. The two posts above do not really address the pruning of minor highways. They just refer to the displaying of traffic jams. If road type no longer factors in then that indeed makes this argument moot. But I don't see where they make mention of it.

The reference speed is so far only being used in the calculation for when the jam coloring is displayed and to what degree for a set of segments. This replaces the associated default speed that each road Type had/has.

Default speed based on Type would only influence routing when there was no historical or real-time data, because the ETAs could be better vs real world until sufficient data was collected.

[bgodette]

Should link to the most useful maps possible; interactive maps.

California: http://earth.dot.ca.gov/
Colorado: http://dtdapps.coloradodot.info/Otis/Flex/MapView
Oregon: https://gis.odot.state.or.us/transgis/

The Florida one is broken like 98% of the time unfortunately. However the data from fdot funclass.zip can be loaded into ArcGIS Explorer or converted to a KML and used in GEarth.

[bgodette]

I agree with you—though the scope of this overhaul was only drivable roads, and really public drivable roads. I intend to revisit the non-drivables sometime after this is committed.

That discussion will likely lead to standardized naming of important navigational points that aren't normal Places, eg "SFO - Rental Return" "SFO - Cell Waiting" "DIA - Arrivals East" etc.