Central MA, US: very odd routing behavior: tale of 2 routes

[icycle]

Hi,
This is a puzzle I face daily and can't solve yet. My commute takes me from the center of Rutland, MA through the center of Holden, MA to a destination in Marlborough, MA. Have a look at the very inefficient route (the 34.5mi one in particular is insane) that Waze generates when I route from Rutland to Marlborough. Keep in mind the insane route goes through the center of Holden.
http://www.waze.com/livemap/?zoom=11&la ... g=65577685

Now let me generate a route from the center of Holden to Marlborough and the result is completely different, and pretty efficient:
http://www.waze.com/livemap/?zoom=11&la ... g=65577685

But watch what happens when I bring the destination closer, suddenly the route from Rutland follows the Holden route to the highway, no insanity:
http://www.waze.com/livemap/?zoom=12&la ... g=22895964

As an aside, while generating the links above, I ran into this completely insane result when I routed from a manually clicked start (at the center of Holden) to a manually clicked end (in Marlborough, MA) that shows two very inefficient route choices:
http://www.waze.com/livemap/?zoom=11&la ... g=65253699

I have performed a LOT of map editing in the area between Holden and the points where I should be picking up the highway in Worcester. All turns and street directionality should be correct--at least they were the last time I edited which was a while ago.

It appears there is something in the routing algorithm that breaks down when the route is longer, since insanity ensues from the farther away starting point, yet corrects when one gets further along that route.

Thoughts welcome, thanks!

[CBenson]

I'm not sure I completely understand the problem. Keep in mind that everytime someone clicks on one of your links, waze recalculates the route based on the current traffic, so the routes you see might not be the same someone else sees. I don't think I'm seeing a completely different result between the first and second links.

One issue to consider is that waze doesn't consider regular streets for routing in the middle of long routes. It seems to me that at least these segments should be made highway segments so that there is some way to get from I-290 to Grove/Main St without using regular street segments.

[harling]

One issue to consider is that waze doesn't consider regular streets for routing in the middle of long routes. It seems to me that at least these segments should be made highway segments so that there is some way to get from I-290 to Grove/Main St without using regular street segments.

Yes! From what I can tell, this would be the proper way to get onto I-190 S, without driving North and making a U-turn--and changing them from Street to Minor Highway (or at least Ramp) will very likely solve this problem. Good eye.

(Making the change is left as an exercise for the OP. )

[icycle]

Thanks for the responses. I think I'll need to paste screenshots instead of permalinks, since the permalinks aren't really permalinks.

The segment you show is not one I would normally ever use to get onto the highway, so I'll also post focused routes and/or screenshots showing the streets I would use.

I'll get to both of these things later tonight.

Thanks again.

[Spil]

I haven't looked into this one yet, but it seems very similar to the situation that paulwelch23 posted about here. Looks like I came up with the same suggestion there that CBenson and harling posted above, so hopefully we're all on the right track with regard to fixing these issues.

[icycle]

Here is a screenshot of a route that clearly shows the preference to minor highways is not always honored, and this is what I can't understand--the point of this post. In this route, Waze took two "street"s (Shrewsbury St in Holden to Doyle St in Holden) to get to I190, this is what I'd call a "good" route:


Yet here is a screenshot of a route that causes me to question Waze's sanity, where it apparently tried so hard to avoid the previously ID'd non-minor highway/ramp segments that it would send me two exits up the highway to make a U-turn and come back down:


To summarize the point here, why was the first screenshot's route allowed to be so efficient, despite the travel on less than minor highways? It seems Waze will sometimes take the most major road possible, even if it's quite a bit longer in both time & distance.

I think I'm starting to understand why I'm getting a sub-ideal route in my normal commute case and this now begs the question of how I choose what is a "minor highway" vs a "primary street". I read http://www.waze.com/wiki/index.php/Road_Types_%28USA%29 and it didn't really clear it up, but I made a number of updates to the map anyway:

• made the segments previously ID'd by CBenson into minor highways
• fixed up the levels that were also messed up in that interchange
• changed some streets to ramps where there were grade changes involved
• changed a number of streets from street to primary street to reflect that they are much higher trafficked than other streets and used as key access streets to the highway.

Result: https://www.waze.com/editor/?zoom=2&lat ... TTTTTTTTFT

Thanks for the help here.

[CBenson]

To summarize the point here, why was the first screenshot's route allowed to be so efficient, despite the travel on less than minor highways? It seems Waze will sometimes take the most major road possible, even if it's quite a bit longer in both time & distance.

My understanding is that waze will use streets in the first (or last) 10 km of the route. (Although I haven't confirmed what the exact distance is). Thus, you get the "good" route using Shrewsbury and Doyle when you start the route in Holden, but not when you start the route in Rutland.

The map looks better to me. Your changes make perfect sense to me.

[icycle]

My understanding is that waze will use streets in the first (or last) 10 km of the route. (Although I haven't confirmed what the exact distance is). Thus, you get the "good" route using Shrewsbury and Doyle when you start the route in Holden, but not when you start the route in Rutland.

The map looks better to me. Your changes make perfect sense to me.

Thanks for the info, learning more and more about the internals now and this helps the frustration level.

So my next question is...should I demote Rte 122A (Main St Holden/Grove St Worcester) from a minor highway to a primary street? Despite it being a numbered route, it feels the same to me as the other access roads. As a minor highway, it's going to take precedence when it falls outside the 10km ends of the route, and that's not really a good thing--sending people down Grove St just sticks them in a mess of lights and one way streets (W Boylston St/Gold Star Blvd in Worcester).

Google Maps seems to show 122A at the same street magnitude as, say, Shrewsbury St or Doyle St, even when I zoom out:
https://maps.google.com/maps?q=map+hold ... 14&vpsrc=6

So I think minor highway is wrong here, but then again, I may be wrong. The other option is to make 122A a minor highway starting at the Shrewsbury St intersection in Holden and north from there, such that the minor highway piece connects Holden with Rutland, but at the Shrewsbury St intersection the choices of how to get to the highway are all considered "equal".

Thanks

[CBenson]

Well I would certainly defer to the local AMs such as Harling on this. I tend to give much weight to the functional classifications made by the DOT. Thus, I would tend to keep anything marked in blue or red on this map a highway. So I would keep Main/Grove a highway. But, I'm not familiar with the area so do what works and makes sense.

[icycle]

The DOT map shows some additional levels of street beyond waze's typical street progression. I.e. they have:

• Interstate
• Principal Arterial
• Rural Minor Arterial or Urban Principal Arterial
• Rural Major Collector or Urban Minor Arterial
• Rural Minor Collector or Urban Collector
• Local

and we have:

• freeway
• maj hwy
• min hwy
• primary street
• street

Thus, a road like Rte 9 (divided, two lanes, but traffic lights) is marked as a minor highway in Waze, and so is Rte 122A (undivided, single lane, traffic lights). In reality, the two are quite different. Therefore, I now think Rte 122A should be collapsed in levels down to a primary street. OR Rte 9 should become a major highway...

Would like to consult with the local AM though, if there is an active one in central MA. BTW, how do I find out the AM for this area? Is the old (Cartouche) map interface with the AM layer still the right way? Maybe I can help out as an AM.