Cycle Travel Question

[Richard Fairhurst]

That's an excellent idea. I've just added some icons to do that - one for 'unpaved', one for 'major road' (=A road or equivalent), one for 'get off and push', one for 'ferry' - so you can quickly scroll down the list and look for the icons appearing on the right.

[mjr]

That's great although I've spotted one oddity already: step 8 of http://cycle.travel/map/journey/13122 is labelled gravel and described as "Continue on St James' Car Park" when I think it's a left turn into a highway=unclassified (and so paved) called Clough Lane. I've looked at OSM and can't see why: can you, please?

[ossie]

I used this brilliant site for a 1500 mile tour through England, down through Holland, Germany to Switzerland on the EV15, and up through France on the EV6. I managed to get over 200 miles from Dorset to Harwich without really hitting a main road which I thought was nigh on impossible...but no!

Of course there was the odd occasion I ended up in a field but that was down to my GPS only able to download 250 trackpoints per route and me being slightly dim when that straight line went between A and C and I really needed B but on the whole it was a fabulous resource..actually better in many places than the officially signposted EV6 route.

5 sets of lithiums, an Etrex Legend and this great little planner....thankyou Richard !

[Richard Fairhurst]

That's great although I've spotted one oddity already: step 8 of http://cycle.travel/map/journey/13122 is labelled gravel and described as "Continue on St James' Car Park" when I think it's a left turn into a highway=unclassified (and so paved) called Clough Lane. I've looked at OSM and can't see why: can you, please?

It looks like it calculated it as St James' Car Park when you originally planned the route, I think - older versions of OSRM were sometimes a bit funny when roads were shared with the edges of parking areas. cycle.travel stores the instructions with your saved journey so it doesn't have to regenerate them every time you look at it (and so your route doesn't suddenly change if I change the weightings!).

If you tweak the route slightly by moving one of the via points, then it'll regenerate it with the latest data and routing code. It looks like the new weightings come up with a different route through the town centre so you may want to tweak it anyway.

That said, it shouldn't be showing as unpaved - that looks like a mismatch in mapping the old route codes to the new ones, and I'll take a look.

I used this brilliant site for a 1500 mile tour through England, down through Holland, Germany to Switzerland on the EV15, and up through France on the EV6. I managed to get over 200 miles from Dorset to Harwich without really hitting a main road which I thought was nigh on impossible...but no!

Of course there was the odd occasion I ended up in a field but that was down to my GPS only able to download 250 trackpoints per route and me being slightly dim when that straight line went between A and C and I really needed B but on the whole it was a fabulous resource..actually better in many places than the officially signposted EV6 route.

5 sets of lithiums, an Etrex Legend and this great little planner....thankyou Richard !

Excellent - really glad to hear it!

[Psamathe]

However it did take me on some dire paths that had me spitting fury and venom and cursing everything from my GPS to the entire French nation.
Sometimes it was due to obvious recent developments, road closures or new builds. Other times, I was exactly following the GPS track and it was an established footpath or farm track but so bad that I could not cycle on it.
One farm track in Flanders left me looking like some of the British visitors to the region 100 years before and I spent a couple of hours hosing myself, panniers and bike down with a graveyard tap.

I appreciate from looking at maps that I was caught in bottle necks between motorways, airports, rivers and canals. So at that time it appears to sink to accepting "anything passable" rather than decent tracks.

Eek - sorry!

Yes, that's mostly the reason - when there's a bottleneck and it can't find its favoured options (minor roads or cycleways). It's almost always because someone has just mapped it in OpenStreetMap as a "track" without any surface information, so cycle.travel has to guess what the surface might be.

Coincidentally I'd actually noticed this problem the other week, and have tweaked the weightings accordingly. So I've just (yesterday) uploaded a new version of the routing database which is much more sceptical about tracks in Europe without any surface information - it now only gives them 40% of the score that it used to give them (basically, it assumes they're dirt rather than gravel). I'd be intrigued to hear what it does if you now try and plan a similar route - hopefully it should be less keen to take you along such a track. But sorry again.

If it helps, when you're planning a route, the blue highlight becomes green for unpaved sections - so you can zoom in on the route, look at Street View pics if there are any (or Geograph in Britain), and drag it away.

(Lots more changes in the new version - the biggest changes in the routing algorithm since the first version. I'll write more about it in a day or two!)

Interesting aspect to routing (not just for cycling but maybe particularly relevant to cycling) is the state of routes. Strikes me that editing the OSM data is something somewhat too complex for many to get into, particularly to just "report" or note a particular issue or inappropriate selected link.

So idea would be for a web page/site to display a zoomable/scrollable map and allow a user to click on a road/track (click on it and maybe highlight start to end of road/track (i.e. between two neighbouring junctions and allow start and end points to be dragged maybe/maybe too complex). Then display a pop-up with a few data entry controls, maybe "Suitability to road bikes" (unsuitable, passable with difficulty, ok, good, excellent), "Suitability for off-road" (same options), maybe include traffic options (e.g. for busy fast roads), etc. and maybe a "Surface" dropdown and an "obstacles" dropdown. Ideally no registration (although requiring an email would be ok) and ideally feeding back into OSM (where possible and options might need adjusting to fit with OSM recorded info). Feeding back into OSM where possible would to me seem better than just using as a local routing site information, though I don't know how OSM get on with input from unregistered users (but it's just "clarification" and "suitability" details rather than significant mapping.

Just a thought - and I appreciate that would involve a fair amount of work.

Ian

[Richard Fairhurst]

It's an interesting area - full of promise but difficult to get right!

A few car navigation apps are now letting their users report issues directly to OSM - Navmii, for example - but there's often a signal-to-noise problem: issues get logged on OSM which aren't actually a problem with the OSM data, but with the navigation app not parsing it correctly.

Or you can do what Mapbox do, which is log the issues internally (rather than on OSM) and then get your squadron of paid editors to make the corrections, but that can result in remote editors stomping on correct local data.

And then there's the new version of maps.me, which lets you edit a few things in OSM (mostly POIs) directly from the app. Pro: thousands more people are contributing to OSM overnight, which has to be good. Con: we've had people using it to persistently add Chinese translations to (say) Italian restaurants in Vienna which don't actually have Chinese names.

I'd very much like to do something along these lines, but without falling into any of those traps. I'm tending towards logging the issues externally (i.e. on cycle.travel rather than directly on OSM), but unlike what Mapbox do, making the list public so any OSMers can view it and fix the issues. Lots to think about...

(Incidentally, if you're not aware, you can now log an issue directly on openstreetmap.org without having to edit the data or indeed register; use the little speech bubble icon on the right. It helps if you do register, because then you've clicked the agreement that basically says "I'm not copying from Google Maps" - we can't guarantee that for unregistered users, so can't always take the reports on trust.)

[Psamathe]

...
I'd very much like to do something along these lines, but without falling into any of those traps. I'm tending towards logging the issues externally (i.e. on cycle.travel rather than directly on OSM), but unlike what Mapbox do, making the list public so any OSMers can view it and fix the issues. Lots to think about...

Maybe (and I can see it being a lot of work), a list of summarised "ratings" and/or "changes" and/or "additions" that presents an "average" of your own user input, as a list with a check box so some OSM person can go down the list checking checkboxes for changes they wist to use then at the end click a "Do It" button which moves the selected changes across to OSM dataset.

Each proposed change could be summarised with a new/old (where available) as well as number of users and range of new data input (e.g. "Un-passable x2, "Bad x3") - giving indication of consistency.

Unselected changes could remain (e.g. until more users have confirmed the input). Maybe even have a threshold number of users needed to use a change (both within your own routing and for OSM to be presented with the change) to improve "reliability.

Such a scheme would probably need to operate on fixed sections of route (i.e. user clicks on track/road and section selected between e.g. junctions or OSM points of some type either side of clicked point and user cannot adjust those points).

Of course one could go on adding functionality forever without any consideration that you/somebody has to sit down and design/implement/test/document/etc. it all.

...
(Incidentally, if you're not aware, you can now log an issue directly on openstreetmap.org without having to edit the data or indeed register; use the little speech bubble icon on the right. It helps if you do register, because then you've clicked the agreement that basically says "I'm not copying from Google Maps" - we can't guarantee that for unregistered users, so can't always take the reports on trust.)

I would have thought that the "I'm not copying from Google Maps" could have been handled by a check box on the submission (at bit like the "I've read the terms and conditions" that is used so often).

I was aware of online openstreetmap.org editing but not that you could do it without registering.

Ian

[SA_SA_SA]

....

I think cycle.travel is very good but as I use its routes as a initial starting point, an undo feature for the last drag or 3 would be useful: is that very difficult?

[mjr]

I think cycle.travel is very good but as I use its routes as a initial starting point, an undo feature for the last drag or 3 would be useful: is that very difficult?

Isn't that what the curly back arrow near the Save and Elevation buttons does?

[Richard Fairhurst]

It does indeed!

I said I'd write a bit more about what's in the new version. There aren't too many visible changes, but a lot has changed under the hood which should result in better routes.

The big change is that it now studies each turn much more rigorously. Previously there was a penalty for right turns, a lesser penalty for left turns, and that was all. (Or the other way round in Europe.) Now, it takes many more factors into account when analysing each turn: what type of roads they are, whether you're turning onto/off a signposted route, and so on.

This has two effects. First, the routes are less "fussy", especially in cities: there's less unnecessary weaving around, and it won't so often suggest you dodge onto a little road only to come back onto the original road 100m later. On one London route I've been experimenting with, it's taken out seven consecutive turns to give a route which is just as good and much less hassle. Second, the turn-by-turn instructions are simpler and clearer, which should be useful for those who like to print out cue points or use a TCX file.

There's also a lot of little weighting adjustments, like the one to be more sceptical about track surfaces in Europe. The under-the-hood adjustments also set the stage for a couple more improvements I've got planned for later this year... but more of that anon.

[SA_SA_SA]

It does indeed!......

Thanks,the undo makes it much more usable.

[Bicycler]

Indeed, a worthy addition. Richard, would it be possible to include the date of the current Openstreetmap data somewhere on the site? It helps people making changes to know whether they have botched up somewhere or whether the data hasn't yet been updated.

[matt_twam_asi]

The 'About Our Maps' page states an aim for at least monthly updates - I'll leave it to Richard to confirm this but it seems about right from manually checking up on edits I've made to OSM. The page also mentions that you should post possible oddities that you've found to the cycle.travel forum.

[Bicycler]

I don't have any problem at all with the rate of updating. I have no desire to bring about Richard's deafening by server noise viewtopic.php?f=16&t=83831&p=956095&hilit=server#p956095 I would just find it helpful to be able to check the date of the current import. A footnote or simple changelog would suffice.

I'm happy to report any problems I encounter with cycle.travel where I know the Openstreetmap data to be fine but wouldn't want to waste Richard's time on those occasions when it's just a matter of waiting for the map to be updated.

[matt_twam_asi]

Sorry if my reply came off as combative Bicycler, I was just mentioning that in the absence of what you're asking for (which would be a great idea) you can probably assume that the data is updated at least monthly. I'll go back to lurking now.