Page 26 of 67

Re: New Route Planner

Posted: 7 Jun 2018, 11:09am
by mjr
From viewtopic.php?f=16&t=122470
Richard Fairhurst wrote:
mjr wrote:http://cycle.travel/map will find many of these, but it is slightly keen on the nodes and sometimes needs a nudge, similar to London (where it's far too keen on the London Cycle Network + IMO).


Slight tangent from Belgium, but I'm always keen to hear of suggestions on this one - London is really difficult because there aren't that many signifiers (based on OSM data) to differentiate a good cycling road from a bad one. I think it's probably the hardest routing challenge anywhere that c.t operates. If you've got ideas, or examples of where c.t is getting it wrong and should do something else, I'm all ears!

I've given an example from Belgium. Examples in London include almost anything involving CS7 and most routes near-but-not-on CS1 or CS8, such as http://cycle.travel/map?from=London+Bla ... ham+Common where I'd suggest CS6 then CS7 is clearly better (easier to follow, shorter, faster) than the route based on NCN 4 and LCN 3, despite the widely varying quality of CS7. It looks like the routing prefers the NCN/LCN combination too strongly over the on-road CSes which seems strange to me, but the CS are currently marked as a Regional Cycle Network - are Regional Routes elsewhere better or worse than Nationals? In Norfolk/Suffolk, we don't have many, but they seem similar quality but more direct. I think most of Somerset's have become three-digit Nationals.

Re: New Route Planner

Posted: 7 Jun 2018, 1:22pm
by Vorpal
I used to go fairly regularly to Colchester Zoo, and Cycle Travel pretty consistently makes different choices in that area than I would.

For example, from Silver End to Colchester Zoo, Cycle Travel puts me on Hollow Road to Kelvedon, then onto the A12 side path, then across the A12 to Easthorpe Road. The openings have been closed on the A12 in this area, but it would, in any case be a horrible, horrible crossing, especially if you had any younger cyclists with. The next crossing I am aware of on the A12 is at Easthorpe Green, abut I still wouldn't consider crossing there, even if I was alone.

A *much* better route is to take the B1023 Inworth Road to New Road, then go through Messing and take minor roads, which is far more pleasant than the A12 side path. One can then come out at Fountain Lane and have relatively short distance on a busy B road, or avoid more motor traffic a bit longer, cross the B road at Birch, take NCR 1, and turn off at a bridleway that goes (almost?) to the zoo.

Coming from Witham, Cycle Travel partly recommends the route I would have picked to go from Silver End.

If I were going from Witham, I would use NCR 16 and NCR 1, then either take Birch Road, or the bridleway. In this case, though, there is nothing wrong with the route recommended by Cycle Travel.

p.s. It is possible to cycle to the zoo from the bridleway that goes north-northwest toward the zoo from NCR1. If the whole route isn't bridleway, or legal to cycle on, I can understand why Cycle Travel would not include it, but I have used it on my bike.

Re: New Route Planner

Posted: 7 Jun 2018, 2:50pm
by Psamathe
Vorpal wrote:I used to go fairly regularly to Colchester Zoo, and Cycle Travel pretty consistently makes different choices in that area than I would.

For example, from Silver End to Colchester Zoo, Cycle Travel puts me on Hollow Road to Kelvedon, then onto the A12 side path, then across the A12 to Easthorpe Road. The openings have been closed on the A12 in this area, but it would, in any case be a horrible, horrible crossing, especially if you had any younger cyclists with. The next crossing I am aware of on the A12 is at Easthorpe Green, abut I still wouldn't consider crossing there, even if I was alone.

A *much* better route is to take the B1023 Inworth Road to New Road, then go through Messing and take minor roads, which is far more pleasant than the A12 side path. One can then come out at Fountain Lane and have relatively short distance on a busy B road, or avoid more motor traffic a bit longer, cross the B road at Birch, take NCR 1, and turn off at a bridleway that goes (almost?) to the zoo.

Coming from Witham, Cycle Travel partly recommends the route I would have picked to go from Silver End.

If I were going from Witham, I would use NCR 16 and NCR 1, then either take Birch Road, or the bridleway. In this case, though, there is nothing wrong with the route recommended by Cycle Travel.

p.s. It is possible to cycle to the zoo from the bridleway that goes north-northwest toward the zoo from NCR1. If the whole route isn't bridleway, or legal to cycle on, I can understand why Cycle Travel would not include it, but I have used it on my bike.

I feel there is a distinction between a horrible crossing (that might take you some time) or using a side path on the A12 compared to Ride With GPS directing along the A140 and then along the A14 (even where for a short distance there is a parallel road they don't bother to take you down).

I can always accept that local knowledge can give better routes in detail but Ride With GPS along the A140 and then along the A14 (major trunk road with all those fast articulated lorries coming out of Felixstowe ...). If they can't even avoid sending you onto such a major nightmare road for cyclists then can they really clam to be offering a route for cyclists?

Of course in practice nobody would follow the Ride With GPS route - they'd arrive at the start of the "certain death" bit and then get out their maps and find a route where they'd still be alive tomorrow. And suddenly the Ride With GPS route they could achieve in the day and on time suddenly becomes a massive detour taking a lot longer.

Ian

Re: New Route Planner

Posted: 7 Jun 2018, 3:15pm
by Vorpal
But where Cycle Travel shows a crossing, there are barriers on the median. I guess one could lift a bike up over the barriers on the median, so it's not physically impossible, but doing so in the face of heavy, 80 mph traffic (with a high volume of HGVs) would be daunting, even for the hardiest of souls. Having driven on and crossed the A14, I might rather cycle on that that try to cross the A12, where there's no crossing. It's a bit 6 of one or half dozen of the other to choose between them. :(

Re: New Route Planner

Posted: 7 Jun 2018, 3:21pm
by Psamathe
Vorpal wrote:But where Cycle Travel shows a crossing, there are barriers on the median. I guess one could lift a bike up over the barriers on the median, so it's not physically impossible, but doing so in the face of heavy, 80 mph traffic (with a high volume of HGVs) would be daunting, even for the hardiest of souls. Having driven on and crossed the A14, I might rather cycle on that that try to cross the A12, where there's no crossing. It's a bit 6 of one or half dozen of the other to choose between them. :(

Is the difference that the OSM maps need updating (you could update the crossing section error) whereas the Ride With GPS is making choices irrespective of the available information about suitability for cycling (i.e. cycle.travel uses the available information to chose the best route for cyclists assuming the OSM information is correct; whereas Ride With GPS is not using available information to chose the best route for cyclists).

So cycle.travel route is maybe a data error (one anybody could correct) whereas Ride With GPS is a systemic shortcoming.

Ian

Re: New Route Planner

Posted: 7 Jun 2018, 3:26pm
by mjr
Vorpal wrote:But where Cycle Travel shows a crossing, there are barriers on the median. I guess one could lift a bike up over the barriers on the median, so it's not physically impossible, but doing so in the face of heavy, 80 mph traffic (with a high volume of HGVs) would be daunting, even for the hardiest of souls. Having driven on and crossed the A14, I might rather cycle on that that try to cross the A12, where there's no crossing. It's a bit 6 of one or half dozen of the other to choose between them. :(

Unless it's been closed since last June and OSM needs updating, it looks to me like there's a gap in the median where you could push a bike across (and cycle.travel does say it's on foot): https://www.instantstreetview.com/@51.8 ... -10.34p,1z

I agree it's not great, but it's not as unreasonable as you're making it sound. Your suggested nicer routes are between 1 and 5 miles longer, so I think c.t is probably correct to use the sidepath and crossing. The basic problem yet again seems to be that Essex County Council and Highways England have been [inappropriate word removed] at cycling transport infrastructure in that area.

Re: New Route Planner

Posted: 7 Jun 2018, 3:38pm
by RickH
Vorpal wrote:But where Cycle Travel shows a crossing, there are barriers on the median. I guess one could lift a bike up over the barriers on the median, so it's not physically impossible, but doing so in the face of heavy, 80 mph traffic (with a high volume of HGVs) would be daunting, even for the hardiest of souls. Having driven on and crossed the A14, I might rather cycle on that that try to cross the A12, where there's no crossing. It's a bit 6 of one or half dozen of the other to choose between them. :(

Google Streetview show a staggered gap (maybe 10m between the access points on either side of the central reservation), about 200m beyond the road junction, where it suggests you cross (the directions call it a footpath). It is where a public footpath ends at the A12 & was presumably installed to allow walkers across. Has that been blocked up or is it still there?

Re: New Route Planner

Posted: 7 Jun 2018, 6:27pm
by Vorpal
My mistake. I looked at Google streetview and didn't find that crossing. I must have started a ltitle too far up the road. There is another, almost identical one at Easthorpe Green, and I thought that was the only one on that stretch of the A12.

It doesn't change the main point, which is that it is far from the best route option.

Why is my preferred route selected from Witham, but not Silver End or Coggeshall?

So, I did a little experiment...
CM77 8ED (The Swan Inn on the A120) to Colchester Zoo
Cycle Travel says 12.8 miles, and 80 minutes and includes the A12 crossing. It's actually a little less than that, as it seems to include some distance around the zoo for some reason, but that's there for all iterations, so I will ignore it. If I adjust the results to my preferred route, it is 13.9 miles and 87 minutes.

On google, they recommend the A120 and the A12 side path, which is 12.5 miles, and the A120, Messing, and Fountain Lane as an alternative at 12.7 miles. If I adjust the results to my preferred route (Cuthedge Lane, Hollow Road, B road, Messing, Fountain Lane) it says 13.3 miles. There is no difference between them in travel time (1 hour 3 minutes), presumeably because the A12 is relatively flat, and the alternatives involve a few hills.

Why use the A12 side path & crossing, and not use the A12 side path to Marks Tey, like Google suggests? Actually, I think that is a better alternative to the crossing, but not as good as my preferred route. I used the A12 side path sometimes when I was in a hurry to get to Marks Tey / Stanway.

I won't say how long it took to get to the zoo with tandem and trailer, and kids who needed bananas opened or something every 5 minutes ;)

Re: New Route Planner

Posted: 7 Jun 2018, 7:47pm
by Richard Fairhurst
mjr wrote:From viewtopic.php?f=16&t=122470
Richard Fairhurst wrote:
mjr wrote:http://cycle.travel/map will find many of these, but it is slightly keen on the nodes and sometimes needs a nudge, similar to London (where it's far too keen on the London Cycle Network + IMO).


Slight tangent from Belgium, but I'm always keen to hear of suggestions on this one - London is really difficult because there aren't that many signifiers (based on OSM data) to differentiate a good cycling road from a bad one. I think it's probably the hardest routing challenge anywhere that c.t operates. If you've got ideas, or examples of where c.t is getting it wrong and should do something else, I'm all ears!

I've given an example from Belgium. Examples in London include almost anything involving CS7 and most routes near-but-not-on CS1 or CS8, such as http://cycle.travel/map?from=London+Bla ... ham+Common where I'd suggest CS6 then CS7 is clearly better (easier to follow, shorter, faster) than the route based on NCN 4 and LCN 3, despite the widely varying quality of CS7. It looks like the routing prefers the NCN/LCN combination too strongly over the on-road CSes which seems strange to me, but the CS are currently marked as a Regional Cycle Network - are Regional Routes elsewhere better or worse than Nationals? In Norfolk/Suffolk, we don't have many, but they seem similar quality but more direct. I think most of Somerset's have become three-digit Nationals.


That's really helpful - thanks.

Finding safe cycle routes through South London is a bit like trying to make a silk purse out of a sow's ear, really - when CS7, usually judged the worst of the Superhighways, is the best option then that illustrates the difficulty! But I'll have a look at rejigging the weighting for the various cycle networks within London. (I need to fiddle with it anyway as neither the CSs nor the Quietways are showing up in the turn-by-turn directions at present.)

Across Britain, my impression is that Regional Routes are almost always country lane routes, but usually with more motor traffic than NCN routes. As you say Sustrans has (outside East Anglia) brought most of them into the NCN. Here in Oxfordshire that's worked fine, but in some other places less so: I think the ex-RR NCN 423 from Usk to Monmouth has too much fast traffic to be an enjoyable NCN route, for example.

Re: New Route Planner

Posted: 7 Jun 2018, 8:40pm
by Rule5
Just like to add my two pence worth.

I gave up on Ride With GPS after being sent down the nth boggy farm track on my lovely new road wheels, with no discrimination between different types of Right of Way. So I was very happy to find that cycle.travel indicates where a route is off road (eg bridlepath) so I can re-route before the ride.

I have been very happy with the cycle.travel route planning when i am in the countryside, and it has introduced me to some infrastructure i didn't know about, eg around Merthyr Tydfil.

However I live in West London, so nearly all my rides begin in that area and I find cycle.travel intensely annoying and virtually unusable because when I am trying to route along well known roads that just get me out of London asap (eg Greenford Road UB6) the planner sends me around every single possible diversion eg housing estates and pointless loops. It takes so long to persuade it that I just want to get along my established routes until I hit the countryside that I have just given up and had to return to RWGPS.

Am I doing something wrong, or is there a fix? I could have a fictional start at the edge of London I suppose but I like to know the distance of the whole ride before I begin.

Re: New Route Planner

Posted: 7 Jun 2018, 9:02pm
by Richard Fairhurst
Vorpal wrote:I used to go fairly regularly to Colchester Zoo, and Cycle Travel pretty consistently makes different choices in that area than I would.

For example, from Silver End to Colchester Zoo, Cycle Travel puts me on Hollow Road to Kelvedon, then onto the A12 side path, then across the A12 to Easthorpe Road. The openings have been closed on the A12 in this area, but it would, in any case be a horrible, horrible crossing, especially if you had any younger cyclists with. The next crossing I am aware of on the A12 is at Easthorpe Green, abut I still wouldn't consider crossing there, even if I was alone.

A *much* better route is to take the B1023 Inworth Road to New Road, then go through Messing and take minor roads, which is far more pleasant than the A12 side path. One can then come out at Fountain Lane and have relatively short distance on a busy B road, or avoid more motor traffic a bit longer, cross the B road at Birch, take NCR 1, and turn off at a bridleway that goes (almost?) to the zoo.


This is actually quite a fun example...!

Like pretty much any route-planner, cycle.travel works by scoring each individual section of road and each turn, adding them up, and choosing the route with the best score. In the example you've given, the two routes (one via the A12 cycleway, the other via Messing) have almost exactly the same score. The difference between the two is just 0.18% (5866 vs 5877, for what it's worth). It's not often that two alternatives are so closely tied.

I think my gut reaction would be to take the Messing route rather than the cycleway route, too. So the question is why c.t is (narrowly) choosing the other route, and there are two reasons that I can see. One is that it doesn't differentiate between a cycleway beside a busy road and a cycleway in the middle of nowhere, because they're both just marked as 'cycleway' in OSM. I'd very much like to de-emphasise this sort of cycleway beside a rural dual carriageway, because they're not generally a whole bunch of fun to ride; and in fact, doing this has been on my to-do list ever since starting the site. But it's a Hard Problem to solve. I return to the problem every few weeks, but I haven't yet found a way of detecting such cycleways that wouldn't take (literally) days of server power...

But the other reason is fixable, and that's to do with the penalty that c.t gives the A12 crossing. At present it knows that it's crossing a green-signed A road, and gives it a penalty accordingly. But such crossings can vary from this, which I do all the time and which is very benign, to the example you've found.

I think I can get the crossing penalty to be higher for the busiest A roads. It will require a bit of jiggery-pokery but it should be doable. That will mean it's still happy to route across the A44 as in my example, but much more wary about crossing the A12. I'll have a play and post here when I've got something working.

Why use the A12 side path & crossing, and not use the A12 side path to Marks Tey, like Google suggests?


In that case it's because carrying on to Marks Tey means you have to follow the A120 for a bit (which has a high traffic count, so c.t will try and avoid it) and go across two roundabouts.

Re: New Route Planner

Posted: 7 Jun 2018, 9:16pm
by Richard Fairhurst
Rule5 wrote:Just like to add my two pence worth.

I gave up on Ride With GPS after being sent down the nth boggy farm track on my lovely new road wheels, with no discrimination between different types of Right of Way. So I was very happy to find that cycle.travel indicates where a route is off road (eg bridlepath) so I can re-route before the ride.

I have been very happy with the cycle.travel route planning when i am in the countryside, and it has introduced me to some infrastructure i didn't know about, eg around Merthyr Tydfil.

However I live in West London, so nearly all my rides begin in that area and I find cycle.travel intensely annoying and virtually unusable because when I am trying to route along well known roads that just get me out of London asap (eg Greenford Road UB6) the planner sends me around every single possible diversion eg housing estates and pointless loops. It takes so long to persuade it that I just want to get along my established routes until I hit the countryside that I have just given up and had to return to RWGPS.

Am I doing something wrong, or is there a fix? I could have a fictional start at the edge of London I suppose but I like to know the distance of the whole ride before I begin.


No fix at present I'm afraid. Essentially cycle.travel's raison d'etre is to avoid roads like this (the first thing I saw when I randomly dragged the StreetView man onto Greenford Road!) and it will indeed try its damnedest to keep you away from them.

I think ultimately the challenge there is that you want two different types of route - a fast route to the edge of London, then a nice route in the countryside - and cycle.travel aims to be a "nice route" planner. Because of the way it works, it's not technically possible to provide route type preferences without renting lots of new servers (there are boring algorithmic reasons for this which I should write down in a FAQ somewhere).

But... as mentioned elsewhere, I am working on a feature that will let you draw straight lines between via points, effectively turning the route-planner off for those sections. It won't give you a route per se (it'll say "head southwest for 500m" rather than "follow Greenford Road for 500m"), but it should fulfil your need to get the mileage right. I've started work on it, and it's not a trivial change to the code, but hopefully it should be ready in a couple of months.

Re: New Route Planner

Posted: 7 Jun 2018, 9:39pm
by RickH
Richard Fairhurst wrote:I think ultimately the challenge there is that you want two different types of route - a fast route to the edge of London, then a nice route in the countryside - and cycle.travel aims to be a "nice route" planner. Because of the way it works, it's not technically possible to provide route type preferences without renting lots of new servers (there are boring algorithmic reasons for this which I should write down in a FAQ somewhere).

I think it still needs a little too much persuading to let the route carry on on busy roads - often taking a right turn (across the busy traffic :? ), a few hundred yards to the end of a block & then turning right back onto the busy road again (once more across the traffic :( ).

As I've said elsewhere I'll sometimes use cycle travel to produce the overall route & then replot it by hand in BikeHike (useful for the OS maps) or RwGPS (easy to share with other folk that I ride with who also use it). That can give me the fast access on a route I know AND the nice route all in one. Plus the hand plotting helps me to know better where I'm going to be going (as in I will think "I go through the village then left after the church", etc.) & how it fits in the bigger geographical picture rather than just being in a little GPS bubble as I follow the line on the map.

Richard Fairhurst wrote:But... as mentioned elsewhere, I am working on a feature that will let you draw straight lines between via points, effectively turning the route-planner off for those sections. It won't give you a route per se (it'll say "head southwest for 500m" rather than "follow Greenford Road for 500m"), but it should fulfil your need to get the mileage right. I've started work on it, and it's not a trivial change to the code, but hopefully it should be ready in a couple of months.

That will be great! Thanks for all the hard work you put into this behind the scenes :D

Re: New Route Planner

Posted: 7 Jun 2018, 10:50pm
by Vorpal
Richard Fairhurst wrote:In that case it's because carrying on to Marks Tey means you have to follow the A120 for a bit (which has a high traffic count, so c.t will try and avoid it) and go across two roundabouts.

What most cyclists do there is cross the A120 (west of the junction), take the side path to Station road, ride through the trees to the signalised crossing on the A120, then use the pedestrian bridge over the A12. I have used the roundabouts there, and they aren't as bad as they look, but the described alternative is much less intimidating.

Re: New Route Planner

Posted: 8 Jun 2018, 6:36pm
by Rule5
Richard Fairhurst wrote:
No fix at present I'm afraid. Essentially cycle.travel's raison d'etre is to avoid roads like this (the first thing I saw when I randomly dragged the StreetView man onto Greenford Road!) and it will indeed try its damnedest to keep you away from them.

I think ultimately the challenge there is that you want two different types of route - a fast route to the edge of London, then a nice route in the countryside - and cycle.travel aims to be a "nice route" planner. Because of the way it works, it's not technically possible to provide route type preferences without renting lots of new servers (there are boring algorithmic reasons for this which I should write down in a FAQ somewhere).

But... as mentioned elsewhere, I am working on a feature that will let you draw straight lines between via points, effectively turning the route-planner off for those sections. It won't give you a route per se (it'll say "head southwest for 500m" rather than "follow Greenford Road for 500m"), but it should fulfil your need to get the mileage right. I've started work on it, and it's not a trivial change to the code, but hopefully it should be ready in a couple of months.


Thanks for your reply Richard. Much appreciated.

BTW Greenford Road can be very quiet on a Sunday morning!