34

I understand that Google Maps works out travel times based on the speed limit for the road and the current traffic on the road. (When selecting the driving function).

When using the cycling/walking function, I assume it works out the travel times by using the average walking pace(~3mph) and the average cycling pace.(~9.6mph)

But does it increase the time if your journey includes a lot of upwards hills, as you would obviously be travelling slower if walking or cycling? And does it decrease the time when there are lots of downwards hills, as you would be travelling quicker (maybe not by foot but definitely on bike)?

Michael Hampton
  • 61,833
  • 6
  • 142
  • 258
JackU
  • 451
  • 1
  • 4
  • 6
  • Google Maps don't have trustable topographical information around the world; maybe the best would be check by yourself –  Jul 11 '19 at 15:38
  • 3
    Answered on [bicycles.se] in this question https://bicycles.stackexchange.com/questions/44793/how-does-google-maps-estimate-cycling-times – AakashM Jul 11 '19 at 15:50
  • @lucasr300 it may not have it everywhere, but it certainly has it in some places, as it shows a graph of the altitude along the way, and includes the total amount of uphill and downhill elevation changes on some (many? most?) routes. – jcaron Jul 11 '19 at 16:16
  • @jcaron Yes, exactly, I mean just this information may not be trustable. If this is important to you, maybe you will need more accurate information. –  Jul 11 '19 at 17:03
  • 3
    If you are looking for elevation-aware routing I’d recommend brouter. Much better for cycling than Google Maps (or any other routing service I’ve tried). – Michael Jul 12 '19 at 05:52
  • @Michael Just had a quicker look at brouter. It seems a lot cleaner and more detailed than Google Maps. Brilliant recommendation. – JackU Jul 12 '19 at 05:56
  • @lucasr300: Assuming they use SRTM and/or ASTER data, it’s 99%: “ASTER GDEM coverage spans from 83 degrees north latitude to 83 degrees south, encompassing 99 percent of Earth's landmass.” Apparently the data is bad for summits over 7km but I guess you won’t use a routing service there ;) – Michael Jul 12 '19 at 07:08
  • 1
    @JackU: You can also install brouter on an Android Smartphone and use it as an (offline) navigation service for osmand (my favorite navigation app for walking/cycling, works offline as well). – Michael Jul 12 '19 at 07:11
  • Google Maps doesn't use speed limits to compute ETA. I have evidence that it often estimated 90km/h average speed over a constant limit of 130. They use traffic averages. – usr-local-ΕΨΗΕΛΩΝ Jul 12 '19 at 08:11
  • I don't know enough to post an answer, but I do know that it completely messes up walking routes in my neighbourhood & tells you the "quickest" way is over a steep hill, even though walking around said hill is far quicker regardless of your normal walking speed. – laszlok Jul 12 '19 at 09:48
  • 3
    If you find that Google is generating steep or otherwise unreasonable cycling routes, you can ask them to fix it. I've done this before, suggesting a nearby flatter route in place of a very steep hill. It takes a few days, but they do incorporate the suggestions eventually. – Kyralessa Jul 12 '19 at 11:29
  • 1
    To the answers, which are suggesting, that it actually does, because they reversed the route: May be it just takes the average speed from the database for that direction. Of course this speed is higher because of the the downhill. So yes, it "does" consider the downhill but it would not be working, when there wouldn't be any historical data from previous rides. – dipa2016 Jul 12 '19 at 11:09
  • This question may be highly dependent on location, and the quality of road/route data and elevation data available to google maps. – Criggie Jul 12 '19 at 13:34
  • It takes everything in to account : *the times are based on (often, millions of) other user's actual journeytimes.* – Fattie Jul 12 '19 at 13:35
  • Google Maps don't just rely on speed limits, it also takes actual travelling time into account. That is, even if the speed limit is 90 km/h on a certain piece of road but it usually is so congested that the average speed is 70 km/h GM calculates with the latter number. – d-b Jul 13 '19 at 13:19

4 Answers4

46

Surprisingly, it appears that the answer is YES, Google Maps does make a distinction between uphill and downhill.

I routed out a couple of bicycle rides from my home (in Germany) to a city either 90 km away (with a 700 m climb) or 60 km away (with a 1000 m climb), depending on the route taken. I then reversed both routes.

In both cases, Google Maps shows a noticeably shorter ride time for the downhill direction than the uphill one.

The 60 km route is 5h1m going up or 4h24m going down. The longer route is listed as 5h34m going up or 4h48m going down.

Of course, Google Maps doesn't know exactly how fast I ride, but I would guess you can take the percentage difference between the uphill and downhill and do some multiplication to figure out your likely actual time based on your average speed.

Kyralessa
  • 3,964
  • 2
  • 24
  • 33
  • It would be interested to understand if the time added on for the uphill is calculated from the angle of the incline or just Google maps being aware that the altitude changes but not of that the incline angle. Obviously if the angle was shallow youd be able to climb 1000m quicker than if it was steep – JackU Jul 11 '19 at 16:02
  • 12
    Google Maps not only consider elevation changes when calculating travel times, but tries at least to a certain extent to avoid elevation changes when using the route planner. If a flat detour is avaialble, my experience is that the route planner often suggest to go around the elevation change instead of up and down, even if the distance is slightly greater. – Tor-Einar Jarnbjo Jul 11 '19 at 18:51
  • 39
    I wonder if feedback from location-enabled devices also factors into the calculation. For driving directions, for instance, Google Maps won't know that a particular road is riddled with potholes, or always has deer frolicking around the embankment, but they will know that a lot of cars are going 5mph under the speed limit instead of 10 over it, and on my anecdotal experience, seem to adjust their time estimates accordingly. – choster Jul 11 '19 at 23:41
  • 7
    @choster IMHO with bikes, such feedback is impractical. With cars, you can generally assume that drivers want to get to the destination in reasonable time and if they happen to be driving slow, they are limited by road conditions. With cyclists, you never know if they are more tired than usual, or just choose to drive slowly and admire the view. – IMil Jul 12 '19 at 00:17
  • 2
    @IMil Another factor you have, (that I don't believe Google takes into account) would be the weather as in wind direction. There is a slight grey area with cycling as like you say there are many external factors that affect the cycle times. – JackU Jul 12 '19 at 05:50
  • @Tor-EinarJarnbjo Ooh, how I wish it did that in Finland. It always picks the biggest hill it can find, even though there is a flat road right next to it in real life, which is just a few hundred meters longer trip. I guess they have accurate topologies only for highly populated areas. – Juha Untinen Jul 12 '19 at 08:40
  • @JuhaUntinen Perhaps Elevation data is indeed not available everywhere, but I have the impression that on a country-by-country basis, data is either available or not available. For Finland, elevation data is certainly there. Google may of course also use other conditions for the route planner, but e.g. in this case, Google suggests the slightly longer, but flatter route via Tollpa instead of the shorter (and faster), but hillier route vie Tottijärvi: https://www.google.de/maps/dir/61.4107129,23.3027702/61.3825784,23.3117385/@61.3949404,23.2796951,13.75z/data=!4m2!4m1!3e1!5m1!1e4 – Tor-Einar Jarnbjo Jul 12 '19 at 09:27
  • 1
    @IMil Taking averages will get rid of most of these biases, I'd say – Jasper Jul 12 '19 at 11:18
  • @Jasper but not of systemic biases. Not far away from where I live, there's a mountain road quite popular with sportsmen. Most of the people taking this road on bike are physically fit quite above average. So, if average times were used in prediction, the mountain road could seem much faster than it would be for your average cyclist. All things considered, calculating time based on slope is more reliable. – IMil Jul 12 '19 at 12:03
  • @IMil In the case of looking at nothing but the slope, I agree. In the real situation, where there lots of factors that may influence the travel time, I'd say that both approaches have their advantages: guessing will probably miss factors all the time, but they may not be too significant most of the time, using actual averages will occasionally have biases. – Jasper Jul 12 '19 at 12:12
  • They don't seem to take the elevation of bridges into account. This is certainly true of the Brooklyn bridge, which is shown at sea level on the elevation profile. – phoog Jul 12 '19 at 13:34
  • 1
    @choster I swear that Google Maps takes me for little detours from time-to-time just because they want data on that particular route. – Yoshiyahu Jul 12 '19 at 19:59
  • 9
    There's no need for Google to perform complex and sophisticated computations. Google has lots of data to work with, elevation may be part of it, but previous history of other travelers is by far the most significant factor. Google doesn't need to know the slope of a hill, all it needs to know is that people going north go slower on this section and people going south go faster. – barbecue Jul 13 '19 at 01:03
  • @IMil By far the bigger effect is that a strong cyclist will take the same road much faster than a weaker one. But none of this really matters: Google can still compute the time taken by an average cyclist, and that's as good an approximation as you're ever going to get. – David Richerby Jul 13 '19 at 08:36
  • @PatrickTrentin The elevation impact on the estimated travel time is so unrealistic, that I think it can be completely ruled out that it is based on real data. I checked a nearby mountain road with quite a lot of bicycle traffic (about 20km distance with 1000m elevation change) and Google claims 2:15h uphill and 1:30h downhill. I have done the trip myself and only 45 minutes difference between the 1000m climb and the 1000m descent can not match anyone's style of driving. – Tor-Einar Jarnbjo Jul 13 '19 at 22:56
33

No. Google Maps works a different way.

Google Maps learns travel times by monitoring the pace of other riders.

The Google Maps app "constantly" sends data about your location back to the Google servers. It knows you're on a bike, not driving, because you requested a bike route, and because your travel time is not an outlier from other people doing the same.

And that's also true for driving; though of course Google gets more data for drivers, so driving data is fresher.

Yes, really. Google "spies on you" and observes your movement, nominally for this exact purpose (but they won't refuse a subpoena in most cases). And yes, you agreed to be a lodestar for others, at some point when you clicked "agree" on a term-of-service while interacting with a Google product. You can deny Maps access to your location, but then of course, it wouldn't work at all.

This tracking is fairly continuous, even when the screen is off, which means Google knows when you stop for a break, and excludes that time from the data.

This provides a very authentic transit time without having to do deep calculation on the effect of grades, tight curves, congestion and road condition on ideal travel times. For instance if a flat, straight rail-trail is heavily overgrown with brush to where you must slow to maneuver around a thicket of overhanging branches, it captures that reality.

Harper - Reinstate Monica
  • 35,577
  • 4
  • 65
  • 144
  • 6
    This is the correct answer. – barbecue Jul 13 '19 at 01:00
  • Google timeline frequently tells me I was driving or motorcycling when I commute. That's definitely a compliment to my riding speed, but I suspect they've got a threshhold value and ignore faster riding, or at least can't tell it from other means of transport. So Goog's idea of cycling speeds are arbitrarily low. – Criggie Jul 13 '19 at 03:08
  • @Criggie that should self-correct eventually if most cyclists do the same. – Harper - Reinstate Monica Jul 13 '19 at 14:27
  • 1
    This is not precisely correct, you have to specifically grant permission for this service. If you go to https://myaccount.google.com/activitycontrols/location you can turn on location history. When you do you will be told this-

    Location History saves where you go with your devices. To save this data, Google regularly obtains location data from your devices. This data is saved even when you aren't using a specific Google service, like Google Maps or Search.

     – deep64blue Jul 13 '19 at 20:54
  • As I already commented on Kyralessa's answer: Looking at Google's estimated travel times on nearby mountain roads with much bicycle traffic, the numbers are so unrealistic, that it is fairly obvious that they are not based on real data. I would assume that even a relatively high number of bicyclists deliver speed data with such a large scatter and distribution of values, that it is difficult to draw any half-way sensible conclusions from the data. – Tor-Einar Jarnbjo Jul 13 '19 at 23:00
  • 1
    Checking another road I know well with a relatively steep climb (400m altitude in less than 10km) on the other hand gives more realistic travel times (60 minutes up, 25 minutes down). There however, bicycling is prohibited, so not only is Google very unlikely to have much data from bicyclists on that road, nor do they seem able to conclude from lack of data that you are not allowed to cycle there. – Tor-Einar Jarnbjo Jul 13 '19 at 23:23
  • 3
    This is interesting information, but I'm not seeing any evidence for it. Could you cite a source? – Kyralessa Jul 14 '19 at 16:51
  • 1
    https://gadgets.ndtv.com/apps/features/how-google-maps-gets-its-remarkably-accurate-real-time-traffic-data-1665385 – bjmc Jul 14 '19 at 17:16
  • @bjmc There is absolutely nothing in the article you are linking to, which confirms that the collection of user movement is used for bicycle route planning. – Tor-Einar Jarnbjo Jul 14 '19 at 19:36
  • @barbecue You have obviously no or very little practical experience with using the Google Maps route planner for bicycles if you believe this to be the correct answer. Yes, Google is collecting and evaluating movement data from most Android users and is also without doubt using this data for car navigation, but as I already pointed out with several examples, for whatever reason and at least usually does not use this approach for bicycle routing. I would guess that the available data is insufficient. – Tor-Einar Jarnbjo Jul 14 '19 at 21:24
  • @Tor-EinarJarnbjo You obviously have no or very little experience with Google cloud services. See how easy it is to make a snotty comment about a stranger? – barbecue Jul 15 '19 at 15:36
  • @barbecue I do. Google Cloud is a runtime environment, in which paying customers can host their 'cloud' (yeah, buzzword) based applications and get access to certain services provided by Google, e.g. big-data storage or machine learning engines. It has absolutely nothing to do with how the route planners in Google Maps work or are implemented. As I have written several times: You only need to check the numbers in a few real examples of bicycle routing, to determine that this is not how the routing works. – Tor-Einar Jarnbjo Jul 15 '19 at 15:45
  • 1
    @Tor-EinarJarnbjo The fact that Google Maps does this with automobiles is beyond dispute, and that makes sense given Google's core competency. I presume you are not challenging that. Let's create the supposition they do not do this with bicycles. That means instead of simply applying their standard model that aligns with their skills, they are conjuring up some new scheme just for bicycles. That doesn't make sense. Why would they do that? The simpler explanation is that they have a fallback in case there isn't enough usable data, but they don't highly maintain that. – Harper - Reinstate Monica Jul 15 '19 at 18:49
  • @Harper Have you read what I already commented? Just three comments further above, I wrote: 'Google ... is also without doubt using this data for car navigation'. Why should you presume that I am challenging that? You also ask 'why would they do that?' I answered that question yesterday and have not yet changed my meaning: 'I would assume that even a relatively high number of bicyclists deliver speed data with such a large scatter and distribution of values, that it is difficult to draw any half-way sensible conclusions from the data.' – Tor-Einar Jarnbjo Jul 15 '19 at 21:55
17

At least for bikes it does (assuming the topographical information mentioned in the comment is available).

I just checked with two cycling routes near my home, both 4,2km long according to Google, one almost flat, the other with a climb towards the end (obviously going downhill the other direction). Result:

  1. 4,2km almost flat but slightly downhill: 14min
  2. 4,2km almost flat but slightly uphill: 15min
  3. 4,2km mostly flat (3km), but a climb towards the end (1,2km): 17min
  4. 4,2km mostly flat (3km) and going downhill at the beginning (1,2km): 12min

If these numbers are of any relevance certainly depends on your bike and your fitness level (and probably some other factors, too). I hate riding uphill and my bike isn't made for it, so whenever I had to "cycle" up the hill at the end of no. 3 I would get off my bike and push it, certainly losing more than two minutes compared to a flat route. The same goes for downhill, I doubt I was ever letting my bike run so fast I would acutally gain two minutes...

Sabine
  • 4,351
  • 1
  • 17
  • 24
4

Short answer: Yes absolutely google maps is aware of elevation changes.

Longer answer: ...but its fairly far from correct.

Example - There's a well-travelled local climb of 2.5 km and 140 metres elevation change. Its an average of 5% but is 10% at the top with a flat bit at the bottom.

Google maps predicts 13 minutes to descend and 18 minutes to climb. That's roughly 40% longer to climb than to descend, which is utterly wrong.

I personally descended in 3:25 and best climb in 10:30, for a 300% increase in climb time vs descent time.

The 20 top riders have descended in under 2:30 and have climbed it in under 6:00 minutes for a 240% difference.

So google maps should be predicting a climb in 39 minutes for a descent of 13 minutes
or a climb of 18 minutes and a descent in 7:30.

tl:dr yes google maps is aware of elevation changes and does try to take that into account, but doesn't fully account for the increased work required to climb a gradient.

Further info:

Walking is predicted to be 40 minutes to climb and 34 to descend.

Driving is predicted to be 3 minutes either way.

Link to route on Google Maps

Strava segment for uphill

Strava segment for downhill

Site is in New Zealand, near Christchurch. I'm roughly in the middle of times for riders on those segments, slower uphill and faster downhill (mass advantage there!)

Criggie
  • 601
  • 8
  • 11
  • 1
    How do your actual times compare to Google's estimates in general? I would guess that their bicycling estimates are probably based on "average" cyclists traveling without time pressure, not enthusiasts trying to get best-possible times. – dwizum Jul 12 '19 at 13:58
  • @dwizum good point - strava's going to be a self-selected group of faster riders who are exercising, and will likely exclude those who just commute for transit. However the difference in predicted time as a percentage is what I was trying to highlight. Using a flat segment at https://www.strava.com/segments/14319538 I get 24 minutes actual vs a google maps prediction of 36 minutes. The best times are under 16 minutes, and the worst are 40-60 minutes (stops mess that up) – Criggie Jul 12 '19 at 14:18
  • 1
    My guess is that google uses the time of people on road racing bikes, much faster than my commuting or even fun ride times. – Willeke Jul 12 '19 at 14:19
  • 2
    I guess all we can really say is, cyclists travel at vastly different speeds, unlike motor vehicle traffic which is generally consistent. I know many casual cyclists who are terrified by quick descents, their % differences would probably be much different than comparing the cyclists you're looking at. This is really interesting data though, it's too bad we can't learn more about the populations in strava vs how Google gets their numbers. – dwizum Jul 12 '19 at 14:23
  • 1
    @dwizum At least in Germany, cars also travel at vastly different speeds... – gerrit Jul 12 '19 at 22:20
  • 1
    @dwizum not super relevant, but you might like https://www.strava.com/heatmap#12.90/172.64301/-43.69822/hot/ride Its a combined heatmap of routes from strava users. Fascinating, but times and segments aren't shown. – Criggie Jul 13 '19 at 02:01