4

Apologies if this isn't directly relevant to cycling, although it does involve a cycling product.

While in Dyserth, Wales I decided to see how my Edge 530 would navigate me home. This is the route it took me on: https://www.strava.com/activities/3812895019 (I've cropped the end for privacy). That is 11km longer and involves twice as much climbing as what Google Maps suggests: https://www.google.co.uk/maps/dir/53.3020881,-3.4178364/53.2006867,-3.2228244/@53.2517539,-3.3319621,12.38z/data=!4m2!4m1!3e1.

I did have popularity routing turned on in the settings and the avoidance settings were set to avoid toll roads (not applicable), unpaved roads (not applicable, unless the map data is bad), ferries (not applicable) and narrow trails (possibly applicable but some of the roads I was directed on are narrower than the more direct route). Major highways were not set to be avoided. The calculation method was set to minimise time.

However, playing around with these settings and setting it to navigate back there from home still seems to suggest the same route, no matter what permutation of avoidance / popularity settings I have (apart from not avoiding unpaved routes, but then it tries to go up a path which is only navigable on MTB and even then is not legal). I've also tried switching between the minimise distance, minimise time and minimise ascent routing settings to no avail.

Am I missing something here? If it's going to suggest routes like that it seems like a useless function.

Update Garmin support is investigating this.

Another update Just tried on another location. Navigating to the Ponderosa Cafe on the A542 / Horseshoe pass with popularity routing turned off and on the different minimise settings:

Interestingly, the minimise time setting actually minimises the ascent and distance as well. But again, none are close to the default Google route which is 23.5km, 356m ascent and 1hr37.

One other thing that is interesting / stupid is that I was actually at that cafe before and thought I'd see how it would direct me home on the minimise ascent setting (but didn't follow it). It tried doing the Moel Arthur route (minimise distance route above, but in reverse).

7
  • Could be map data, but I doubt that as the route it took you is longer than alternates with no common sections, unless there are two errors. Therefore most likely an algorithm error. Routing algorithms are well understood, but complex. Unlike google maps that has near infinite resources in a server farm, small devices need find results with limited computing resources and a highly optimized. One optimization is 'giving up' on a potential route, which if it happens too soon, means the route is not considered. Report it to Garmin (when they get their servers back up).
    – mattnz
    Jul 26, 2020 at 2:25
  • @mattnz I had wondered about that possibility but it seems like it would have to be exceptionally poor at routing given that near Trefnant, it put me on an A road and could have taken me back entirely from there on A roads, albeit still not as directly as Google, so it's hard to see what would make it give up there. At the very least I would think that a basic routing algorithm would rank A roads as faster than B and unclassified roads. I also have a Garmin car satnav so I'll try that just for comparison. Jul 26, 2020 at 3:00
  • I’d recommend preparing your routes at home on a big screen. For example brouter ends up with this route for your trip: brouter.de/brouter-web/#map=12/53.2585/-3.2846/…
    – Michael
    Jul 26, 2020 at 7:10
  • @Michael I could do, but it seems to take the spontaneity out of rides if I have to plan them in meticulous detail beforehand. I cycled to Dyserth on a whim from another location I'd ridden to first (the routing from there to Dyserth was fine) and then wanted a reasonable route home. Jul 26, 2020 at 15:43
  • 1
    @Criggie Not yet. I wonder if they can't be bothered fixing or it's too big a problem as I've had an email asking for feedback but no resolution at all. I've also sent them details of the Ponderosa issue. Aug 14, 2020 at 1:36

2 Answers 2

5

My guess is that "popularity routing" is weighted quite heavily in the choice of route, and "minimise time" is a secondary weighting.

I see Google has routed you on the A541, which is definitely more direct, but may be less "compatible" with cyclists and therefore less popular. The Strava route has used B5429 which is even narrower, but seems to have less traffic, based on Street View.

Strava and presumably Garmin can access their pool of data about where people ride. I'd posit that the climb out of Y Groes Fawr is rather popular, having 200 metres hard climb after a 100 metre of more-gentle climb. That's the kind of effort where pacing is critical, and it sorts the rabbits from the tortoises (the go-hard-and-blowers from the pacers) Specifically https://www.strava.com/segments/6671074 the Moel Arthur climb.

Putting that to the test - here's strava's heat map showing both google and your routes.

enter image description here

Visually you can see that the Strava route highlighted in Blue is brighter than the other three google routes, so while people do ride all those paths, the one you rode is more travelled by Strava users.
My first version of this had bright magenta for the google highlight, and it seriously obscured the point. The green and blue are chosen to both have similar low luminosity levels. So ignore any stray hot-pink dots please.

Search around more at https://www.strava.com/heatmap#12.14/-3.32479/53.25998/bluered/ride and see how the results match your personal experiences - the good riding roads should be more-ridden than the horrid roads.

3
  • 2
    But OP says that they played around with various settings, including popularity routing. I think it’s more likely to be a bad (or highly “optimized”) routing algorithm which due to some heuristics ends up going southwards and west of the mountain range.
    – Michael
    Jul 26, 2020 at 7:09
  • 2
    Yes, I've tried turning off popularity rating and it does the same thing. I've just tried again from home to navigate to Dyserth and it gives me the exact same route, both with minimise distance and minimise ascent. The route was especially annoying from Dyserth though, as the Moel Arhur climb from that site is particularly gruelling: 20% in sections. I walked up most of it. Maybe I'm a particularly weak cyclist, but it's not the type of hill I want to just be casually directed up if it's miles out of the way. Jul 26, 2020 at 15:41
  • 1
    @CuthbertAnnihilator fair point. Perhaps the next step is to contact Garmin support about "sub-optimal routing" and see where they can take it. Either way - you have some loverly-looking roads for riding, I've got nothing quite like that and I''m envious :-) !
    – Criggie
    Jul 26, 2020 at 19:21
1

Some routing algorithms really don't like UK A or even B roads, and will do silly things to try to avoid them. A short stretch of dual carriageway to allow overtaking is another potential feature for the algorithm to avoid. Many services use openstreetmap data, and tags there can be wrong, so it's possible there are one or two places the Garmin doesn't think riding is possible (and of course it uses old data as it's working offline)

I've had a look, using Komoot for playing. This uses OSM data and comes up with something similar to Google's first choice, though (as is its habit, especially set to "touring") avoids the A451 more than Google, adding 40m of climbing and 800m distance. A reasonable route anyway. It can't be that the Garmin was trying to avoid A roads, because of what it does near Trefnant.

It's strange, because a lot of algorithms go out of their way to put you on cycle routes, but NCR5 is offered by Google and not Garmin.

I can't see how a single-point bug can cause the route you got, because of where it goes after Trefnant.

Two partial suggestions, neither of which I can prove:

  • It's doing something similar to what some early car satnavs used to: Navigate to a nearby known point, then use a library of known routes from there to a point near your destination, then navigate properly again. That would be really dumb in a bike computer, but it's possible that disproportionately popular routes can have a similar effect, because even if I force a route via Denbigh (a plausible anchor point for that old method) I still don't get the massive climb, just lots of A road.

  • It doesn't think you can get over the A55 near Bryngwyn Bach airfield. I've seen this problem in the past with roads mis-tagged as one way (that was Google maps) or bike route bridges mis-tagged as footpath bridges (that was in OSM data). That alone wouldn't explain such a circuitous route, as evidenced by Google's other two routes.

Strava's popularity (which I used to use a lot before their route planner became a paid feature) would sometimes pull the route in from a long way away; round here that often meant popular commuting routes, i.e. direct but busy, the opposite of your problem. Strava have a huge dataset; I wonder how big Garmin's is, or was when it was downloaded to your device. We of course don't know their weightings, but it's possible that every segment you rode had a history of a few riders, while some segments on the more direct route had none at all.

With all those options, I'd expect a "minimise climbing" option - it would be more relevant to bike routing than some of what you did have.

BTW if one option to avoid is literally called "narrow trails" I would read that as avoiding singletrack and maybe some silly bike paths, i.e. if you can get a car down it, it's not a narrow trail.

7
  • 1
    I know this is a bit rambling. In the end I posted it before I added even more vague thoughts.
    – Chris H
    Aug 11, 2020 at 19:15
  • There is a minimise climbing option (called "minimise ascent") but it makes no difference. There are three minimise options - ascent, time and distance. None of them makes any difference. At the moment I have a ticket open with Garmin, but I suspect that it's too big a job to fix quickly as they've already sent me a feedback survey. Aug 13, 2020 at 14:56
  • That fact the the "minimise-" options make no difference really suggests a bug, and I suspect in the software rather than the map data for the reasons I give above, though obviously it works well enough, most of the time, for most people. Whether the options make exactly no difference or no practical difference (e.g. "minimise ascent" changes the route from 500m of climbing to 490m, when you could do in in 200m) would be of interest to anyone trying to track down the bug, but wouldn't help users
    – Chris H
    Aug 13, 2020 at 15:15
  • Komoot's routing options are generally okay, but I find it tends to overestimate actual paved surfaces. I can nearly count on it that when Komoot thinks it is "paved" it's actually "gravel" or "dirt". Manually changing the route is quite easy though and the export options fairly friendly so I've kept to it.
    – user33335
    Aug 14, 2020 at 5:01
  • 1
    @gktscrk that comes directly from openstreetmap data. I've fixed a few local issues, so can you. It's possible one or more editors has been rather optimistic in your area
    – Chris H
    Aug 14, 2020 at 7:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.