What's neat here is how clean their elevation data seems to be. Having spent a long time trying to work with various dataset like the NASA SRTM and USGS files, I quickly came to realize how hard it is to clean up this data. Looking at this feature I tried out a couple of things that would have tripped up a naive implementation and Google Maps seems to have dodged the most obvious pitfalls. For example a bike route that crosses a freeway in a trench correctly shows the route without discontinuities. It also seems like Google are getting around some of the problems just by being intentionally vague and blurry. If you make a route short enough it will just say "mostly flat" instead of trying to draw a misleadingly detailed profile.
They're almost certainly using barometer data from Android devices to help know elevations and differences between roads. The barometers can measure altitude differences as small as ~0.5m - it's pretty amazing.
That's interesting. What kind of datastream do you get from that sensor? Something that has a useful ∆Z over short intervals but essentially no absolute accuracy?
You can check it out yourself if you like - I provide a live feed of barometric data from Android sensors at http://pressurenet.io/. The data I collect is polled every 10 minutes, and while the exact pressure is often nearly useless due to noise, trends in the data are very clear. I'm looking specifically at the weather, so over periods of ~6 hours we get really really clear trend lines.
When a device moves rapidly in a short period of time, you know that the sensor data will be reflecting the altitude changes and not the weather changes. Moving the device from my feet to my head shows a clear change of ~.5mb typically.
it looks clean, but from the few routes i checked it might be clean because it's been fuzzed beyond the point of usefulness. it's got no scale on it so i can't be sure what exactly is being showed, but there's a couple major hills near my house (and i mean major, i live in the mountains) that don't even register.
Google already has lots of people working on Maps to make the data better, correct mistakes, etc. They almost certainly have data about bridges and tunnels they can use here to make sure that you get a straight line for them instead of following the terrain profile.
Heh, Google is catching up. I want to highly recommend https://www.komoot.de/plan for hike or bike planning. It's fundamentally based on OSM and very very nice.
Google simply doesn't have the data in the UK, it has probably a tenth of the footpath and cyclepath routes compared to OSM, let alone all the information about surfaces, unofficial paths etc.
Share the recommendation for komoot, although it's quite new, and a bit buggy.
OSMAnd is also useful, not so polished, but you get a lot of power over the way you use the map, for instance searching for attractions, pubs, cycle parking, etc along the way, and it's completely offline.
Sometimes I've heard people bitch about the US-centricness of the Internet and how it can be difficult to enter structured data (like phone numbers or addresses) in a format that isn't compatible with US standards.
It's interesting to see that problem from the other direction. That site is useless for me because it can't easily decode US addresses.
http://transportdirect.info has had gradient profiles in its cycle planner for years. I should know, I was involved in writing it (the gradient profiler). The gradients come from official OS data. It is somewhat slower than Google to calculate routes, but for my money has slightly clearer instructions and seems to make better use of cycle paths.
As someone who really enjoys cycling I have been surprised at how bad mapping and routing applications are, google maps have large portions of the world that are unavailable offline, its impossible (or hard to find out how_ to save routes offline, regularly sends me down dual carriageways and motorways. Open cycle routes loses your position every time you rotate the device and generally lacks quality.
Am I missing something? with the amount of effort gone into OSM I dont really understand why there isnt a top quality app to display and use all that data.
I think they must be using sensor data from GMaps. Yesterday I mapped a ride that said mostly flat. It was not after riding it with Google Maps giving me directions while on the route. Today it shows an incline of 504ft. Same destination, same origin just two different days.
Six years ago, I spent an hour learning to drive manual transmission in my friends car. I did not get very good in that hour.
Then I went and bought a manual transmission car of my own. The drive back home from the dealer was one hell of a trip. I learned very quickly how tricky it is to get a car moving when you're stopped on a hill.
Oh man. I learned to drive in my parents' automatic mid-sixties Impala, so when when I bought my first car - a yellow 1969 Toyota Corolla - I wanted an automatic. The salesman said, "You know, this is just a two-speed automatic, and it doesn't perform very well. I really think you'd be happier with the stick shift." I told him I didn't know how to drive a stick shift and he said, "That's no problem, I'll teach you."
So we drove a few lurchy blocks around San Francisco - yes, San Francisco, the city of seven hills. Ah, if it were only seven!
After a while he said, "Doing great, drive me back to the dealership and you're on your own."
I knew my way home, but I couldn't remember where the hills were.
I could have used elevation data that day...
"...sweat beading upon his brow, he might recount that Damp Morning when he drove his Manual Transmission up the Impossible Grade, and was forced to stop, just below the top! In frantic pantomime, he’ll pull the emergency brake and disengage the clutch. Crane his neck to peer anxiously at the car sniffing his downhill bumper. Bulge his eyes. Gun the engine. Pop the clutch. Release the brakes. Lay down some rubber with a piercing squeal. Float his steed slowly onto the flat. Wave the smoke from his eyes. Pump his arms in brief celebration. And finally, grouse about that sadistic driving instructor who got him into the pickle in the first place. What a jerk!"
Yes, the hand brake is your friend when starting on hills, but it's tough when you're starting out because it means you have to coordinate three things at once :)
Yeah, bought my first car from a town about an hour from my home, a manual car. On the drive home I learned to drive a manual. That weekend I learned to change a clutch :-)
What bike routes need is some indication as to whether a given cycle route is actually any good. 'Traffic levels' for bikes is what I want. Some of the proper routes with little blue signs are great for ten year old's but no good if you are on a reasonable commute.
I followed Google Maps suggested bike route and the ones provided by Transport for London for several weeks before I figured out that I should use the main road in certain sections. Plenty of other cyclists were on the main road going roughly the same direction. The 'cyclist' route had me going on quieter roads that were just not that good. I would find myself going against the main flow of the traffic at junctions and having to navigate a lot more of them, never with the lights in my favour. Plus the distance was longer.
Why is nobody on Strava on this road? Is probably the best way to decide if a bike route is any good.
Well, at least Strava now has a global heatmap: http://labs.strava.com/heatmap. Makes it easier to avoid roads less travelled, I guess.
Regarding the data quality: At least in Germany the ADFC maintains its own cycle path network which also contains metadata about road surface, motor traffic and such things. This is also one of the sources Google licensed for their bike stuff in Maps (although back when they introduced it they had a very hard time matching up their two or three different data sets and there were lots of errors where junctions wouldn't connect, etc.).
I use google maps bicycling directions every day on my bike trip across the US and just saw this the other day.. that plus street view to see how good shoulders + road conditions are can be very cool tools for planning a bike trip.
Sadly in west Texas good shoulders tend to be covered in chipseal resulting in days/weeks of much vibration.
Needs some tuning of the mostly flat definition. When a route contains roads where you're looking at several 100s of meters > 10%, it is not considered mostly flat by the average cyclist. On the contrary.
This is a really nifty addition, I can't wait until it shows up in the mobile app. I recently went on a very hilly bike ride and it was neat to find out exactly how steep/high the hills were.
