kentsmith9 wrote:What am I missing?

It seems to me that it is not universal that the street for the destination will be drivable. If the address is for a street that is not driveable then waze wants to route the driver to the end of the non-drivable street.

Petervdveen has given these examples in the past of addresses associated with non drivable streets:

http://goo.gl/maps/FA1De

http://goo.gl/maps/Qcv6f

http://goo.gl/maps/gNTXk

So I guess the first question becomes - how should addresses on non-driveable streets be mapped in waze? It would be fine to map them and associate addresses with them and then route the end of the segment (or the nearest locations to the drivable road network) if there were no other walking trails on the map. The subsequent question then becomes - why do we have walking trails that have no destinations on them mapped in the US?

I agree. This was a change that was implemented with no notice to the community and that significantly impacts routing to locations near currently mapped walking trails. I totally agree that the change appears to be a solution for the future which has been implemented in a system that is not prepared for it.

I also agree that we need to either 1) get waze to change the way it treats walking trails for routing purposes or 2) change the way we map walking trails (i.e. delete recreational walking trails).

I don't think we have been told that the ability to route to a GPS coordinate cannot take into account the road name when trying to find the closest street. In fact, I thought making this happen was one of the primary reason that waze has not switched to using its own house numbers for address searches. However, this is not a simple rule to implement. There are plenty of addresses that are not closest to the street that bears the name of the address. Examples that come up frequently are businesses that have parking lot roads between the building and the street, in which case waze should route to the parking lot road rather than the street with the address name. The examples that I frequently cite are the buildings that have vanity Pennsylvania Ave addresses in Washington DC even though there are other closer streets.

We have been having a similar discussion in the Texas forum here:

viewtopic.php?f=237&t=64267

We have been having this discussion over in the Texas forum. We have seen repeated problems in Dallas and Austin with long walking trails in green belts adjacent to major streets.

Are you sure about Waze not routing down the walking trail? I know it will try to route you down a pedestrian boardwalk, as it did that several times last month after someone connected the boardwalk in Venice, CA to the road system.

Is the routing behavior different if there's no walking trail drawn in?

Thanks for the Summary CBenson. I guess what I am struggling with is the possibility of Waze sending you down places it should not if the penalties of other roads are higher to that end it would be interesting to know what those penalties are. Not sure if anyone has a "simple chart" or something like it...

8 Lane uncongested highway = 0 penalty
.
.
.
driving through a median with a red arrow onto train tracks at elevation level 10 (or whatever worse case scenario you can imagine) = 5656362456 penalty?

...what are those penalty numbers? what are the values in between?

Can the penalty of a walking trail be made so astronomical that it would never be used (2^32 or 2^64 or something like it) ?

I am coming from another routing failure thread that ran into this problem. Thanks to CBenson for pointing this out to me, otherwise I would have never expected a walking trail would have caused so many unique problems in routing in a group of condominiums.

If I understand correctly from what CBenson has described, Waze intends to have walking trails (and possibly other non-drivable segments) be a part of routing to an address to help people get to their final destination.

It seems like this objective will be better met when Waze uses internal addresses and allows stop points for an address to be something other than the named street address (like a driveway or parking lot road).

In the mean time the current routing method is including a walking trail that often is not connected to a road and then sometimes forces navigation to end at some completely unrelated street leaving the driver completely confused.

It would seem the best solution today with external GPS locations would be to have the routing engine use the street name in the destination to find the spot on that named road that is closest to the destination. I understand it may be the back of the building, but at least it is the correct road name and anyone could simply drive along that road until they find the front entrance. This is certainly better than being routed to a road the ends at some walkway that may or may not have any way of getting a person to the destination.

What am I missing?

But in the case I was running into (link above)

1. All of the addresses were on the same drivable street
2. Most of the addresses were routing correctly
3. Only a few addresses were physically closer to the walking trail so Waze tried to route to them
4. Because we are not on our internal addresses, Waze is using the GPS coords to the closest road without regard to the road name (fail and I believe easily fixable)
5. That walking trail segment was unroutable, so Waze defaulted to the closest segment to the end of the walking trail (sending people almost a mile from where they needed to be and no visibility to the road they need)

End result is completely normal addresses and drivable streets are rendered unreachable by navigation because of the poor selection of the Walking Trail near the destination with no name associated with that destination.

So it seems we have a good solution for the future for some countries (ability to route to walking trails when the address is actually on the walking trail), but it is being implemented in a system that is not prepared for it (causing routing problems on normal streets).

If Waze cannot route to the street name listed in the destination, for the US I propose we delete all walking trails that appear to be closer than the street for any destinations to prevent these errors.

Have we been told that the ability to route to a GPS coordinate cannot take into account the road name when trying to find the closest street? As a past programmer this seems ludicrous. I hope that is not the case, but I don't see it in Bugzilla so I added it last week. For those with voting ability, please be sure to raise visibility to this problem. https://bugs.waze.com/show_bug.cgi?id=6291

I think even after internal addresses makes it we still need this feature. When internal addressing does not have a specific address being requested, I understand that it tries to find the two closest addresses to the one being searched and then estimates along that road where that address would likely be located. I know other Navigation system use that method. Therefore in those cases the routing would need to do the GPS estimation along the named street in the destination.

[/rant]

CBenson wrote:I don't think we have been told that the ability to route to a GPS coordinate cannot take into account the road name when trying to find the closest street. In fact, I thought making this happen was one of the primary reason that waze has not switched to using its own house numbers for address searches. However, this is not a simple rule to implement. There are plenty of addresses that are not closest to the street that bears the name of the address. Examples that come up frequently are businesses that have parking lot roads between the building and the street, in which case waze should route to the parking lot road rather than the street with the address name. The examples that I frequently cite are the buildings that have vanity Pennsylvania Ave addresses in Washington DC even though there are other closer streets.

I completely understand there are drawbacks to the basic solution of including the street name, however until we implement internal address with stop points, including the street name as the final routing address when we only have a GPS point has to be better than routing to a residential road behind a department store that is closer as the crow flies, but in fact might be 2-3 drivable miles from the entrance of the store. At that point you cannot even see the department store, let alone navigate. If I get left on the main street in front of the department store, at least I can figure out how to get into the parking lot.

I guarantee my proposal (include street name as final destination for GPS only routing) would have less problems that what we have today.