vectorspace wrote:Exploring another topic, I wonder if txemt and Driving79 can expand on their observations of how neighborhoods marked with Private Roads do not route well. This would help us understand how to address this in parallel with understanding the issue. We might also update the Private Installations page.

This morning, I asked txemt the following (in an off-forum convo), and he thought your input would be useful. Since you've brought it up here, I'll ask it again, slightly edited. Please let me know if this needs clarification:

As to the nature of the out-routing issues you've found when areas are set to all-Private: Is it a matter of Waze choosing the wrong exit, but using a path through the complex which would be direct, but just not optimal, because it can't tell the difference between better collector roads and lesser residential roads? (Instead of routing to closest exit, trying to route a straighter course, through the installation, perhaps through non optimal roads)

Or is it just random weirdness, where the routing seems confused and wacky?

HavanaDay wrote:I also defer to driving79 or txemt because as said before I have been trying to prove this theory wrong for a while now but everytime I come up with something to test it doesn't seem to route out correctly.

I have no particular angle in pursuing this, other than understanding what the causes are. For this reason, I would like to be given a specific example of "not routing out correctly", so that I can mentally process what might at the root of it.

Fredo, there is never any need to set the entrance streets to private and the other streets to PLR. They are effectively the same thing. You either set everything to private, or set the entering segments to private and the rest to regular street. Note that the latter also has the advantage that if there is a PLR within the community, that you can gain normal PLR through avoidance with this setup, whereas if all the streets are PLR or private, Waze will treat the true PLRs as Just Another Street in the Grid.

There is one thing that we can do to do a partial private implementation. Have a brief split in the Matheson road at Sawtelle. The incoming lane would be private, the outgoing lane would not. This accomplishes what private installations are supposed to, except only in this direction on this one street. Waze will not route traffic onto it, bit it will route traffic out of it.

It would be a tiny flattened diamond, or two half bowties if you will.

It will still have the downside of local destinations being unnecessarily routed around this street, but at least it won't affect any other routes in or through the neighborhood.

CBenson, I don't think they are referring to through routing. Rather, they're talking about problems with internal routing and routing out.

There's been some buzz o the "Routing and TTS" Google Hangout that indicates things are broken, but not enough specifics for me to comment further. Nick, Otto, HavanaDay know more.

That looks good, but the first two sentences can probably be shortened a little bit, perhaps along the lines of:

As of August 2014 the Waze client displays Private Roads identically to public roads. Since this may change in future releases, never use the Private Road type for unrestricted public roads.

As to SV and Sat shifts, I did once check on this. SV pictures are taken with GPS coords, and I found they generally matched up with what Waze records for a drive. You can repeat the test as follows:

Find a road where the Sat imagery is off from road geometry, and appears to be due to a Sat image shift

Drive the road. At each intersection, turn in sharply, pause for a few seconds, make a K turn, return to the main road, and continue.

When the drive shows up in your Waze drives, note all your K-turns showing little squiggles on the drive record. matches

Use SV to trace the path, and see whether SV 1) matches your path pretty closely (the blue dot goes to the recorded coordinates of the picture), and 2) SV should show each intersection at the point where you dived in and K turned.

There's occasional error in the SV coordinates, but they do usually track very well, and seem less subject to warping than Sat pics.

I'm afraid I still don't understand why Waze is doing this. From what we know, the private road penalty occurs at the transition to public road. In Gazzo4U's example, whether you go south and turn onto Hickory or go north and turn onto Square Lake, there is no penalty until you reach one of those two roads, and then there should be an equal penalty turning onto either them.... in which case, the penalty is a common factor, and won't affect the relative outcomes of the two routes, and it should choose the true shorter or faster route.

if we were to implement the proposed solution, the same thing should happen -- equal transition penalties no matter which way you go, since they will all have the same exit configuration.

I'm not saying the proposed solution doesn't work, I just don't understand both why the problem occurs to begin with (meaning, we're missing something in the algorithm or we've found a bug), and similarly why that is eliminated by the split connection.

I did have one thought. Perhaps, especially for shorter routes, the penalty is actually the majority of the trip cost. Say I have a "normal" routing cost of 1 minute for north and 4 minutes for south. But, if say the transition penalty is 100 minutes, that becomes 101 minutes versus 104 minutes. Perhaps waze considers that <3% difference negligible and chooses randomly.

But I don't think that's right either.

CBenson,

My puzzlement is that the same exact holds true both south and north, so there's equivalent no penalty or no penalty for both routes in his example.

Also, I thought you guys brought back from dev that the penalty is on the transition LEAVING the private road, not ENTERING the private road. Or is it only when leaving after having entered?

OK, great, I remember seeing that somewhere, forgot about it, thanks. It might not be a bug, might be a decision they made to sort of force "quick exits" from the private roads where possible. Or might be a bug Even if intentional, it has a bad assumption, which is that if you are on the same segment, you are close to a particular entrance, and certainly has unintended consequences for such cases.

If that's the case, one could indeed resolve it along the lines of your proposal, by simply having an "unnecessary" 5m segment at each junction with public roads.