New route-planner!

[fuaran]

For something like the Lairig Ghru, it would be useful to tag it with mtb:scale http://wiki.openstreetmap.org/wiki/Key:mtb:scale
and maybe sac_scale http://wiki.openstreetmap.org/wiki/Key:sac_scale

Then you know if its mtb:scale of 2 or more, its probably best avoided, unless you are on a mountain bike and suitably skilled. Or willing to push/carry the bike.

[Richard Fairhurst]

Just fixed it such that it won't send you via Lairig Ghru any more; it chooses the northern route via Nethy Bridge. Route-finding in rural Scotland should now be a whole bunch better than previously thanks to a healthy sprinkling of dark magick (or, alternatively, Redis, Lua and industrial quantities of spreadsheet-mangling).

[Richard Fairhurst]

...aaand here's the explanation of what it's doing: http://cycle.travel/news/avoid_hgvs_with_our_new_upgraded_route_planner . In brief: it now uses real traffic data for A roads, so it knows the difference between a quiet rural road in Scotland and a congested street in London with hundreds of HGVs. Next challenge is to get traffic data incorporated for B roads too!

[tom_e]

Nice - I'd wondered how easy such a thing might be.

Have you considered using such a method to tweak the map rendering?  I'm always a bit cynical of computer routing and tend to like eyeballing the map also.  If the traffic levels were somehow reflected in the road rendering (without overwhelming the map with a heavy overlay) that might be interesting.

[andrew_s]

As a possible routing enhancement, would it be feasible to allow a "don't use this section" as an alternative to the drag and drop reroute, possibly by using a right-click?

It likes to follow NCN 41 eastwards from Tewkesbury, but the section between Aston on Carrant and Beckford goes through an unsurfaced section in a field that gets pretty boggy. It would be nice to say "don't use that" rather than "do use" some alternative.

[Richard Fairhurst]

Oh, I know that path - or "ploughed field" might be more accurate. It's a bit devil/deep blue sea though; the alternative is the A46 which isn't exactly a bundle of laughs either. OSRM (the backend routing engine I'm using) doesn't have an "avoid" option, though I wouldn't have thought it'd be impossible... will ponder.

tom_e - the map rendering is a really interesting idea. I've always liked Michelin maps which do that, though from a driver's point of view: faster dual carriageways are shown differently from others, and so on. It wouldn't be trivial - matching traffic data to the OSM source data isn't 100% precise. That doesn't really matter for routing (if a 100m stretch isn't matched, it's short enough that it won't impact the penalty much), but it would look daft on a map. But I'll put it on the long-term to-do!

[bumper]

Is it just me, or is the routing getting worse?

Every time I plan a route it keeps finding crappy little tracks that are unusable. Also if I click on a section of road then click on it a few hundred yards head, it sends me down side roads and back out instead of keeping on that section.

Two local tracks that it try's to send me down are 100% impassable. They're so overgrown the trees have claimed the route.

Is there a way to stop it from routing off road and on cycle tracks?

[andyoxon]

Minor thing, but just wondering how often teh planner updates for 'find photos' from Geograph website, I have some ~20 images recently uploaded, but they don't appear as associated with relevant routes.

[Panoramix]

I've just tested it. Bizarrely I get routes more to my liking in France than in the UK.

I think that it is too "A road shy" in the UK. For instance in Bristol using the A431 as an alternative to the cyclepath to go to Bitton is OK (relatively slow A road). Unsurprisingly the planner will offer to use the cyclepath but if you try to drag the route to go through Hanham (A431), it will insit on using really minor streets to avoid the A431 at all cost.

[Richard Fairhurst]

bumper: can you give some examples (e.g. the two tracks you mention)? That usually happens if the tracks aren't fully tagged in OSM - i.e. there's no surface specified - but I can take a look.

andyoxon: whenever I ask Barry from Geograph for a copy. Right now it's on an ad-hoc basis but I'm hoping that it can move to updates (say) every couple of months.

Panoramix: getting the A-road routing right with the data available is a bit of an ongoing challenge! The A431 has 22,000+ cars a day, which is high - similar to the A5 through Milton Keynes or the A449 north of Worcester, neither of which I'd like to cycle. But if it has a 30mph limit that should ameliorate it slightly. I've just looked and the speed limit isn't tagged in OSM; it ought to be. (French road classifications in OSM are more closely aligned to the general cyclability of a road than British ones are, so it's easier there.)

[Panoramix]

Panoramix: getting the A-road routing right with the data available is a bit of an ongoing challenge! The A431 has 22,000+ cars a day, which is high - similar to the A5 through Milton Keynes or the A449 north of Worcester, neither of which I'd like to cycle. But if it has a 30mph limit that should ameliorate it slightly. I've just looked and the speed limit isn't tagged in OSM; it ought to be. (French road classifications in OSM are more closely aligned to the general cyclability of a road than British ones are, so it's easier there.)

I can understand why it is so hard.

For the French road classification sadly that's less and less the case as they are declassifying N roads (A equivalent) into D roads (B equivalent) while the traffic still stay the same. N means state managed and D means managed by the local authority. It used to be that the state wouldn't trust local authorities but now that money is scarce they are changing their point of view.

You probably already have thought about it but sorting out roads depending on streets that people actually use to cycle on would help a lot. For that you would need data from a strava like website aimed at utility cyclists. That probably is a mammoth task...

[Richard Fairhurst]

Happily, the French OSM community don't faithfully mirror the road classification in their tagging - they choose the levels of OSM tags according to the road's perceived importance in the highway network. So a D road might be the top level (highway=trunk) or the fourth level (highway=tertiary).

That means I can rely more closely on the OSM tags, preferring the lower levels, than I can in the UK. Over here, by contrast, green-signed A roads always have highway=trunk, B roads always have highway=secondary, and so on. To take a really extreme example, that means the A14 or the A1 has the same tagging as the A887. Using traffic data is how I'm trying to fix that, but it's a gradual process.

I'm already using cycling numbers a little - they come as part of the traffic data, and basically I tone down the penalty for really busy A roads that also have lots of cyclists. (The A431 has a fairly low number of bikes: 218 versus 22,000 vehicles.) There's a few companies trying to collect this sort of data through smartphone apps, but unfortunately they're keeping the data to themselves rather than offering it openly OSM-style - figuring that it has some resale value, I guess. Even then, I don't think anyone yet has the critical mass of data except Strava, and as you suggest their data really isn't targeted to utility cyclists - a look at the Strava heatmap for Cambridge, for example, shows high numbers on some busy arterial roads, yet low numbers on popular cycling cut-throughs like Garrett Hostel Lane.

[bumper]

Richard,

These are just two examples that are local to me that the router uses. There are numerous ones I've encountered whilst route planning. I know you can't know which are impassable but I'd help if there was an option to use the roads only, especially when riding 23c on a carbon road bike. It also keeps trying to deviate off the highway onto side roads and dirt tracks, even when the road is straight and traffic free.

53.0224,-2.2667

53.0079,-2.3191

 

[Richard Fairhurst]

Oh lawdy, that's some confused OSM data right there. The route around the old mine (?) is tagged as a cycleway with a "dirt; gravel" surface (just like that, including a semicolon)... and because cycle.travel doesn't recognise that weird combination, it defaults to the standard surface for a cycleway, which is nice smooth tarmac. Which is why it's sending you that way.

Looking at the aerial photos, it should be tagged as a track (at best) rather than a cycleway, and have a single surface tag. But I'm loth to change it myself, not least because the aerial photo is clearly out of date and doesn't show most of the new housing estate. Don't suppose you fancy editing OSM to make it right...?

(In the meantime I've got an update going live early next week which should give more accurate results on tracks/footpaths of any type. I can't do a "no roads" option, I'm afraid - it'd basically require renting a second server - but I think the update will fix a few issues, OSM data permitting.)

[Panoramix]

I'm already using cycling numbers a little - they come as part of the traffic data, and basically I tone down the penalty for really busy A roads that also have lots of cyclists. (The A431 has a fairly low number of bikes: 218 versus 22,000 vehicles.)

In my view the A431 should certainly not be offered as the first option but it can be useful to go from A to B (I've been using it daily for 2 or 3 years). The proportion is low but 218 is still quite a few people in absolute value. That would be a lot of local cyclists on a death wish!

In this light may be when your data shows that some cyclists use it regularly a road should be upgraded from "don't use at all" to "avoid if possible"?

I just feel that slow A roads can be OK in some instances.

[Richard Fairhurst]

Oh, absolutely. A lot of this actually comes down to how the route-planning engine treats turns. If you've dragged the route onto the A431, chances are you want to go on the A431, not be diverted off it immediately. The way to do that is by having a penalty for turns (and that's good, no-one wants a really wiggly route) but the turn penalty code in OSRM, the underlying engine, is quite simple at present. I know exactly how I'm going to upgrade it but it's going to require some scary C++ which isn't exactly my forte...

[Panoramix]

[...] going to require some scary C++ which isn't exactly my forte...

Your C++ is infinitely better than mine!

[andrew_s]

Oh, I know that path - or "ploughed field" might be more accurate. It's a bit devil/deep blue sea though; the alternative is the A46 which isn't exactly a bundle of laughs either. OSRM (the backend routing engine I'm using) doesn't have an "avoid" option, though I wouldn't have thought it'd be impossible... will ponder.

Currently, I'd just drag onto the road through Kemerton, but that's a significant enough extra distance that it leaves me wondering if a barrier wouldn't result in a completely different route that you wouldn't spot on a drag & drop (eg going off round Greet/Gretton to the north of Cleeve Hill)

[andyoxon]

Richard, for some reason the planner thinks Chipping Norton, Oxfordshire  is in continental Europe...       A general issue with the planner going on I think.

Sorry - we can't plan a route between the British Isles and Europe (there's a moat in the way). Please plan journeys to/from your ferry ports.

[neilrj]

Richard, for some reason the planner thinks Chipping Norton, Oxfordshire  is in continental Europe...       A general issue with the planner going on I think.

Sorry - we can't plan a route between the British Isles and Europe (there's a moat in the way). Please plan journeys to/from your ferry ports.

I've just had this too ! I tried to get Talacre to Flint (both Wales).

Never been able to fault this website before, and it isn't really a fault when "Flint, Wales" is used - and anyway "there's a moat" made me smile!

[Ningishzidda]

Richard, for some reason the planner thinks Chipping Norton, Oxfordshire  is in continental Europe...       A general issue with the planner going on I think.

Jeremy Clarkson live in Chipping Norton, and he's a cuntinental.

[Vince]

May be its related to the Tom-Tom routing algorithm which thinks the best way to get from Bristol to Dover East docks is via the Channel Tunnel AND the Calais - Dover ferry.

[neilrj]

May be its related to the Tom-Tom routing algorithm which thinks the best way to get from Bristol to Dover East docks is via the Channel Tunnel AND the Calais - Dover ferry.

Don't dis the Tom-Tom routing algorithm you'll have ningish~whatever coming over all faint

[Ningishzidda]

A TomTom route from Bristol to Dover, Marine Parade, by bicycle is 195 miles through central London, with an estimated duration of 15 hrs 41 minutes.
No tunnel or steam ferry.
The alternative, is 200 miles, 16 hrs 09 mins, via Basingstoke. No tunnels or ferries on this one either.

Googlemaps cycling route is 222 miles, 19 hrs 22 mins. Walking, its 188 or 192 miles, but I can’t be bothered to check them for bicycle accessibility throughout their lengths.

I'd go the 195 miler. 

[Vince]

I specifically stated East Docks in Dover.... makes a difference.