Why routing alogrithms from different maps don't always give the same route

Yeah, this is my biggest bugbear.

I don't know how it is in other countries but it's a big issue for me in the UK as we have a large network of little country roads that are frequently single (or 1.5) lane and poor visibility. I've had a few journeys in areas I didn't know and Google maps took me down lots of twisty country roads to save a (theorised) couple of minutes/miles compared with sticking to the simpler main road.

I've seen a "minimize turns" option in some navigation app.

Also, during heavy traffic, I'd be fine with making 30 turns and continuously recalculated rerouting if it saved significant time.

Yes! Apart from the standard "fastest" and "shortest" options and your suggestion, I'd also love:

* Scenic routes (i.e. navigate as far away from buildings and as close as possible to water, forest)

* Least polluting routes (i.e. least amount of traffic lights, most time spent in the optimal "speed vs. pollution" of your car)

This, but for biking and without a knob.

Make an econometric model of my previous recorded routes with explaining variables: elevation, distance driven already, time driven already, time of day, temperature, road quality, wind direction, rain and explained variable: speed for the next kilometer.

Predict my speed on potential routes based on that.

Calculate routes taking these predictions into account.

I think this may already exist?

My car's built-in navigation presents me with the most straightforward route by default, and shows me options when they approach. If I take one of those options, it has a personalisation setting that will prefer that option the next time.

I'd settle for maps telling me if a road is paved or not. As a motorcycle rider that likes to take less well traveled roads, I've gotten into bad situations due to not knowing a road was unpaved until I got there. Not the least of which was having to be rescued in the middle of the night in freezing temperatures because I had to back track over 20 miles after a paved road turned to gravel half way to my destination.

This would be such cool feature. :) Google Maps often suggest me routes where I wonder if I'll end up in someone's backyard.

I'm not sure there is all that much to be inferred simply from the fact that I am in an area I am familiar with, as each person's preferences are to some extent both idiosyncratic and mood-dependent. For example, there are places I usually avoid around sunset on a clear day, because experience has told me that, at certain times of the year, sun glare leads to bad driving in these areas. I simply might not generate enough data for this pattern to be inferred, even with complete astronomical and meteorological data for the area.

I like to review the proposed route before starting out (and when on the road if I am not the driver), and if I don't like the options presented, I sometimes add an intermediate destination to influence the route-finding, even though I have no intention of stopping there.

Seriously. I experienced this in Gatlinburg last year when I tried to follow the GPS route and it took me up this twisty mountain passage. I’m not sure vertical distance was factored in but my guess is that I went 3-4 times as far.

On the flip side, when we do road trips with my wife, if we have the time we often turn on "avoid highways" and take a more scenic route. Highways can be boring. But I agree, the map should take into account preferences.

Apple Maps does this for me, when cycling - 'fastest, simpler' routes are displayed, along with options to avoid hills

Google maps is exceptionally limited to the point where it’s very easy to beat on mapping features so there is a lot of room for competitors to survive in various niches.

- The offline experience is garbage. Default behavior should cache large routes offline immediately when I’m on WiFi.

- partial connectivity experience is garbage. Stalled map loads, freezes, etc while driving through a 1x cell service area, even if map was pre-downloaded.

- ads for places I don’t care about.

- unable to automatically show places I do care about.

- searching for exits sucks

- searching for rest areas sucks

- measuring distances requires interrupting the navigation

- no way to view current speed, elevation

- filtering for restaurants with drive throughs very frequently ignores the “with drive through” part of the query.

- gas prices come and go on gas station display. No way to filter on price.

- why in the world can’t we filter based on restroom cleanliness?

- trailer towing friendly parking lots?

- enter max range between drives, give me fuel stops optimizing for gas cost or cadence of stops. Mapquest did this in the early 2000s FFS and you printed out the result.

- overlay recent breadcrumb history on map without it getting blown away any time navigation is interrupted.

- search map without interrupting navigation in a way that doesn’t assume something should be added as a waypoint. (Sometimes people use phones for more than one thing at once on a 12 hour road trip.)

The list goes on and on. There are like 4 use-cases that work with google maps and they are all for basic users and nothing else.

Routing is a pretty deep problem. I don't know whether Google actually puts all that much effort into it. For example I prefer to use bbbike.de for cycling routes in my area because it usually provides better routes than Google, even though it's maintained by just one person.

What google did with maps is amazing, but for me a lot of my friends it's becoming a second choice. Coverage with data in my area is impressive, traffic jams info is amazing, but in general algorithms became way "too smart".

With google maps I am being dragged through streets that barely have any asphalt because someone with a tractor went through there. I have a tiny road nearby that is more of a sidewalk for locals, goes tightly between buildings. Google proposes it as a route for a few years now, but only Uber drivers try that - most cars have to fold mirrors to get through. But if someone tries, google treats it as a success and proposes it for everyone. It's painful because all the data about that road and surroundings is there, and when in offline mode it would never let me drive through there.

A lot of friends went for smaller map suppliers, I'm using Osmand~ on longer trips, or that new TomTom app "amigo", as these work like maps used to work 10 years ago - prioritize for typical routes, not the crazy ones. If Osmand had a database with speed cams as big and dynamic as the competition I would never look back. But now I just use two maps - one for route, one for speedcams and warnings.

Yeah, the company has moved further from the consumer market (the devices you talk about), and is now looking much more at B2B markets, providing maps, traffic, routing etc to car companies for their integrated systems. Somewhere on its website it also talks a lot about enterprise so ride hailing services stuff like that.

I personally prefer having the navigator stuck in my cars CD player permanently. No need to pull phone out and try to connect it or mount it each time. Just turn on the car and navigator starts.

I use TomTom GPS on my phone and the user interface/map view is so much better than google maps. Waze could be a possibility, but requires a constant mobile data signal.

Tomtom is selling data to businesses that require raw map data to build their own apps like ride haling service, delivery service, etc..

It's astounding how bad satnav apps are in general. First of all, google maintains a monopoly on apps that work with Android Auto. Using either of them, you'll inevitably come along a route it's guiding you down which is impassible for some reason (roadworks, too narrow, etc)

That's not really a problem in itself, it can't be omniscient. The problem is that there's no way of dealing with it. You turn off, to take the only route available... and the apps reroute you to the same road you can't take. You then have to manually navigate until it gets the hint and routes you down a passable road.

Maps is also heavily dependent on a stable internet connection, strobing and not letting you start a journey until it's connected. Caching your local map is a manual step you have to do repeatedly.

If you stop for fuel & disconnect your phone while you go to pay (as the station might recommend) then you might find it has discarded your journey by the time you get back with no path to re-joining it. Fine if you're going to a single destination and have a good internet connection. Not fine if it's multiple points or - as tends to happen in the middle of nowhere - you can't connect to the internet.

Multiple points are a problem in themselves though. The apps support destination-based routing but are completely foreign to the concept of journey-based routing. In 2022, if you want to drive a scenic route along a given road, your best bet is to take a notepad and jot down the path yourself.

And of course the voice assistants are broken. Garbage. It's absurd we convinced ourselves they stopped being awful.

I'm getting a different experience. Google seems to have been sending me on a slightly longer route to avoid lefts where there are no stop signs for incoming traffic/traffic lights.

I stopped using Google Maps a while ago because it gave impossible directions too often. Use Waze mostly now.

Maybe Google Maps just didn't have enough data about that road or there was a jam on the highway. There are a lot of factors to consider and knobs to tune. In old days navigating software was just looking for a shortest possible way, which given that only main roads were known, was quite a reasonable algorithm. Nowadays all the main players have not just highways and city streets, but also trails and unpaved roads, traffic information, road danger (bad pavement, potholes), et al. There are quite funny accidents because of all these, e.g. I remember reading about Google Maps sending people via a muddy road where most of the cars stucked [1]

1. https://edition.cnn.com/2019/06/26/us/google-maps-detour-col...

>I still don't understand why the curvy road was the default option.

Because when the locals punch in an address across the mountains and it says highway they ignore it and take the shortcut so the ML learns the shortcut but not the "except in blizzard conditions" exception.

The article is about solving a A-to-B routing problem, for which you would use something like a Dijkstra algorithm, which is not a heuristic (in its standard settings) and has nothing to do with solving a TSP.