Long Distance Route...again

Hi all,

I have a UR where a user is driving from Wagga Wagga to Windellama. The issue is they believe the route from Goulburn should go down Windellama Rd but Waze is taking the minor highway to Tarago and then up.

The issue might be that in July Lumley Rd and Cullulla Rd were upgraded to Minor Highway to fix another long distance routing issue (according to map comment) although I note from other sources that they definitely seem like only Primary Streets. This means Waze see a minor highway all the way until just south of Windellama and so is choosing to go that way.

Options seem to be to try reverting Lumley and Cullulla to PS OR elevate Windellama to MH.

Thoughts?
Gav…

Overview link: https://www.waze.com/editor/?env=row&lon=149.72486&lat=-34.97574&s=1428157911&zoom=0&mapUpdateRequest=13795494

Yes. Personally I will often do a manual route recalculation to see if there is a better option but your run of the mill user will just want it to work.
After Breadalbane Rd overpass might also be the distance point where the algorithm is happy to use a PS vs a MH given the short(er) distances involved.

Vestigal did the upgrade so hopefully he will chime in (I PMed them as well)

Its such a slippery slope upgrading all these roads as we will end up with everything being a minor highway. I wish Waze HQ would adjust the algorithm for long distances or provide some more options for the user (such as “I’m happy to use lower category roads if its faster”)

Sorry for the late reply mate.

He and you absolutely right. Just to summarise with links for anyone else reading, Waze should route him from Goulburn down Windellama Rd to Windellama, not the long way along Braidwood Rd. This is confirmed by checking Wagga to Windellama on Google maps, and from Goulburn to Windellama - the long route is only offered when the origin is far-away like Wagga - suggesting the Waze long-distance routing issue.

Even the route times between live map and WME Route Speed are different. https://imgur.com/RJaCYOM

I don’t know what the answer is though. We promote these country roads to primary streets to improve routing, but that’s not enough, so we upgrade some to minor highway, but then we we need to upgrade the PS to mH too now? I don’t know where it ends.

This isn’t a helpful post upon review. Perhaps senior global input would be helpful? There’s not really any other countries with a road network like Australia though.

Yes agree with all of that. What to do though. Tell the reporter sorry we can’t fix it and do a recalc half way?

Tamar is speaking to the routing team for us. Let’s see what comes back.

If there is a possibility that Waze HQ will change the routing engine should we keep a list of roads we artificially elevate so we can regress the change if necessary?

Also when you are talking to them can you point out that sometimes the “correct” route is shown as an option but #2 or #3 but they are faster. If Waze just gave the fastest route as #1 that, I think, would help a lot.

I’ve mentioned it before, upgrading roads in this way causes these types of flow on problems. We need to agree to a proper hierarchy (based on intended road use) and stick to it. Let the Waze algorithm work, stop trying to overcome it.

If is is not working then that is up to Waze to fix, not us. We only map the roads, we don’t (and shouldn’t) do the routing. If we keep “fixing”. how can Waze find the “right” fix?

I agree with you but you are assuming Waze is trying to find the right fix.

ituajr thanks for the tip I didn’t know we could search map comments. Perhaps throw that in the wiki (use of “promoted”) so everyone knows that is the standard?

.
As long as it takes. If the users don’t report it as well, Waze can’t track how big the problem is. Same for if we keep “fixing” it for them.

We don’t know how to fix it at all. We guess and we assume. That is all that we can do because we don’t know how exactly the Waze algorithm works (and rightfully never will).

We’re close but still need to iron out some ambiguity to overcome personal preference and bias. Terms like “as needed to make long-distance routing work” are way too ambiguous.

If we upgrade a section based on how we think it should be routed, how is Waze supposed to use that as an example to fix the problem? We need to be able to show that it is a bigger problem than Waze think and we can’t do that if we keep changing things. If it takes 5 years then so be it, that is on Waze as the app developer, not us as map editors.

I’m with age4670 in that we shouldn’t be trying to fix Waze routing when we have no direct input. I understand ituajr’s perspective that we should do something since Waze don’t appear to be doing anything in this regard.

For me, this prompts the question of how can we effectively get user feedback to Waze? What will Waze users that receive excessively long routes and otherwise far from optimal routes do? Will they; put up with it as they don’t know any better, or become discouraged in using Waze as a GPS for their driving? I do my little bit to promote Waze to friends, but they do become discouraged at some of Waze’s oddities.

Part of that discouragement is that there is no effective “route was crap” feedback. There are some ill-defined “general error” options that land on us map editors, but eliciting more insightful feedback with only meagre parts of the Waze route and driver route taken is difficult. I don’t know even if we could elicit some useful feedback from the driver we can do anything apart from tinkering with the map.

We are all uniquely identified by Waze, so the data should be available, but we map editors don’t have access to it, and it doesn’t appear that the Waze folk do much analysis of it.

Firstly, I agree with all I haven’t quoted from your post. In regards to the above the answer is that we are supposed to bring our concerns up with our Global Champs. They then have a network with the other Global Champs and Waze HQ. This is how I understand our line of communication should work.

I have suggested previously that we have a wiki entry for long standing problems so we don’t have to keep wondering what is going on (although you will wonder why it is taking so long). Whilst I don’t anticipate anything being solved in this regard anytime soon I will admit that when I used this process we got the “Av vs Ave” TTS problem solved and got approval to use the “island fix” to rectify the coastal water boundary issue.

Sorry all for the delayed response. I reached out to Tamar as discussed, she did discuss it with the routing team but the outcome was “unfortunately the routing algorithm will not be review or changed in the near future.” It’s a known issue and it affects many countries, not just us.

The workarounds encouraged was to 1) set the route as preferred in WME, and 2) continue to send them problematic routes (origin and destination) “any time you find one, and I will keep showing them to the routing team. This is what they always ask, examples, and the more the merrier.”

The problem with setting the route as preferred is it locks lower editors out from editing it and Australia doesn’t have a lot of higher level editors. Therefore, I feel like continuing to upgrade routes to primary streets and minor highways is probably the preferable method. I think we might need to come up with a revised and agreed upon strategy for upgrading rural roads and place it on the wiki so we can approach this problem systematically as a country.

Thanks AJ, agreed.

Some points to consider:

  • Most rural roads aren’t intended to be residential streets, but alternate routes to destinations.
  • We have to consider our handling of unpaved roads.
  • We need to ensure we do not compromise the existing road hierarchies which work in non-rural areas and as per design.
  • This may involve us deciding to have a separate hierarchy for non-rural road networks. If so, we’ll need to clearly define where it makes sense to have it applied.

I haven’t worked on a lot of long distance pruning problems, but here’s a first attempt for people to build on:

Introduction

  • Before starting, you need to first read the article on how pruning works to understand how long distance routing works.
  • Waze routing algorithm aims to get drivers on higher level road types as quickly as possible for long distance routing and will ignore lower level road types which may be faster mid-journey. We currently map a rural roads as a street or primary road type, and therefore, Waze long distance routing send drivers longer distances/times despite a short-cut being available mid-route.
  • In order for the Waze routing algorithm to see the short-cuts mid-journey, we need to make the roads more attractive based on how important they are in a driver’s journey.
  • All editors need to consider we are mapping a road network and not just roads. Consideration for how drivers will get between destinations need to be considered when a road is created, updated or reviewed.
  • In rural Australia, “road quality” or a designation, or lack thereof, by a road authority does not accurately predict travel time on that road.
  • In rural Australia, a “destination” is typically located within clusters of other destinations, such as towns, cities or regions, with little in-between.
  • Mapping within clusters should be the subject of Australia’s traditional road type selection. The importance of roads on routes in-between clusters need to be given more importance to be considered by the Waze routing algorithm.
  • We need to shift our focus from “what quality is this road” / “we can only use x road type if an authority classes the road as y” to “how important is this road in routing between destinations.”
  • Higher level road types should never just “stop”, but to continue on, always linking to other clusters, thus forming a road network.
  • The proposed alternative road hierarchy isn’t intended inside a cluster (city/town/region). Such as routing from Burwood, NSW to Granville, NSW, but for routes like Burwood, NSW to Balmoral, NSW 2571 or Barmedman, NSW to Thuddungra, NSW. We need to step outside our comfort zone and consider how we would classify the more obscure routes!
    Possible Road Categorisation

The focus of these road type definitions is for connecting clusters (towns, suburbs, regions) which each other.

Freeway - Locked at level 5 - Should remain exclusively for dual-carriage way roads such as motorways, freeways and express ways.

Major Highway - Locked at level 3 - As a guide, an authority designates the road as a major highway or road, or is named with alphanumeric numbering starting with A, B or C (and possibly D for detour routes, they have to be pretty good, right?).

By assessing roads in the greater area, a “major highway” would stand out as the highest capacity road in the area except for “freeway” road type AND is intended to link with other clusters. If the road terminates into a cul de sac or dead-end, or a network of roads which do not connect with the rest of the road network, it is not suitable for “major highway”.

If the road is unpaved, it can only be allocated “major highway” if a road authority designates it so, or a CM or higher designates it essential for long distance routing such as in outback highways.

If I zoom out on max on WME, I have 50km visible. Is it worth setting an aim for a major highway every 100km? More input needed.

Minor Highway - Locked at level 2 - The most common road type used in connecting clusters together. If the road is unpaved, it can only be allocated “minor highway” if a CM or higher designates it essential for long distance routing.

Primary Street - Locked at auto / level 1 - Use of this road type should be kept to a minimum in rural Australia. It will be most suitable for unpaved roads that connect clusters of destinations.

What does this look like?

Consider two cities/clusters next to each other in rural Australia. The highest capacity / fastest road between them should be the “major highway”, all others should be considered “minor highways”, and any unpaved routes no higher than primary roads.

Where the purpose of the road changes from linking clusters to linking to something else, like giving access to residential or industrial areas, fire trails, dirt tracks, etc. it is more appropriately a “street” with the “unpaved” attribute flagged where appropriate.

I’ve written this as a draft to spark ideas. I’m not suggesting we implement this as is. This is not complete. Build on it if you want. Point out problems and provide solutions. If you take nothing else away, please consider the concept of linking clusters with any other suggestions. :slight_smile:

Let’s solve long distance routing in Australia!

I think there’s a lot of good thoughts there - but it probably merits being moved into a discussion somewhere in Aus rather than NSW.

Agreed. I also think we might have to start a trial area (eg. central NSW which I look after (as it has the most issues of this type I am aware of))

Looking at the Barmedman to Thuddungra example, I see the routing algorithm working exactly as intended. Finding the fastest route while still preferring minor highway over primary street.

The road types for the direct east-west route look to be what they should be. But, the speed looks to be stuck on the default low due to either no one with Waze using those roads or a lack of mobile coverage. This is what looks to be causing the hugely inflated times (as well as some segments on one of the alternatives that have corrupted data).

Even upgrading the road via Bribbaree to minor highway may not be enough for this to become the optimal route (I haven’t done the maths to see whether it would be enough to offer it as an alternative).

The problem then becomes not a Waze routing or data problem as such, but one of the Australian environment (e.g. lack of mobile reception and/or Waze use). How do we approach this?

The idea of clusters fits in with the hierarchical road network design and is worth exploring, but I think making the majority of roads minor highway is a step backwards for the future.

PS Have we ever received an answer specifically on who sets and can change default average speeds or is that part of the routing algorithm? Is it country based?

There was a thread where the default speeds were posted but I can’t find it unfortunately. The highway ones from memory are WAY too slow. Try the Global Champs to see if it is possible to change them.

Here is another UR. Forbes to Trundle in NSW. https://www.waze.com/livemap/directions?overlay=false&to=place.ChIJk5KIBCjWGmsR0HVDkLQJBgQ&from=place.ChIJb4IvPOSMGmsR0HlDkLQJBgQ

Thoughts on elvating some of the primary streets to MH to fix this. Specifically “Bogan Way” and “The Bogan Way”

https://www.waze.com/editor/?env=row&lon=147.81291&lat=-33.10440&zoom=4&segments=397219006
https://www.waze.com/editor/?env=row&lon=147.90023&lat=-33.12555&zoom=4&segments=364390018

At this stage I would say no. Two reasons:

  1. Waze is receiving data (see the GPS tracks). Telstra and Optus indicates mobile coverage along the entire length so that shouldn’t be the issue. Having said that, route speeds both confirms and conflicts with this.
  2. Waze has said they are introducing in the next month the ability to raise the default speed limit. This would be a good example to see the difference.

From Bogan Gate to Trundle I don’t see any problem with the current routing.