Categorisation of Key:smoothness values

[rhhs (Richard)]

While considering the discussion of the images in the smoothness Gallery, I realised there may be 2 conflicting ways of categorising the smoothness key values.

Before I added the “descriptions” column to the Values table in https://wiki.openstreetmap.org/wiki/Key:smoothness in early 2021 (discussion mostly on the talk page here), there were only some often inconsistent photos to illustrate the various values. My reasoning behind the descriptions is as follows: the “anchor” of categorisation should be the difference between the bad and very_bad categories: bad is usable by “robust wheels”, while very_bad is usable by “high clearance”. My interpretation of this is that very_bad is characterised by the presence of obstacles for which high clearance is needed, while bad doesn’t have such obstacles.
For highways that are better than very_bad, 4 categories are then available. From my own experience as a car driver, there are 3 kinds of road qualities: roads that are “no problem”, roads that are “starting to be a problem” and roads that are “a real problem”. They are characterised by the speed you can use on these roads: “full speed”, “careful speed” (still making good speed, but with increased attention and ready to press the brake pedal), and “snail’s pace” (slow down considerably because there’s real risk for damage to the suspension system, tyres, etc. but not the underside of the car). For the last one, bad can be tagged, intermediate for the intermediate one, while the “no problem” either good or excellent can be tagged (there is no practical difference between them for cars). good and excellent can then be used to make a distinction for small wheeled vehicles like skateboards and roller skates.

The makers of the gallery at Berlin/Verkehrswende/smoothness - OpenStreetMap Wiki (translation here) appear to have used a different categorisation where very_bad isn’t as bad as what I had in mind, while I would hardly slow down for the surfaces shown in the bad column. The table has been developed for urban use, where high clearance is hardly ever needed, so maybe that’s why the photos in the very_bad column don’t show obstacles for which high clearance is needed (except for the very_bad asphalt one, maybe). @Supaplex030 you uploaded most of those photos, can you enlighten us on the reasoning behind the categorisation? Or anyone else who took part in its development?

The smoothness Gallery is using many photos from the Berlin Verkehrswende table, which were then used for the smoothness quest in StreetComplete.

I would like to start a discussion on how to resolve the divergence between the two categorisation approaches.

[Supaplex030]

The Berlin table was the first step towards making smoothness a more objectively mappable and reliably analyzable construct. At the center of our considerations was the question of how to set the evaluation standard for smoothness in such a way that, on the one hand, it does fit the existing definition according to vehicle/tire types, but on the other hand also offers sufficient differentiation scope for the evaluation of urban road qualities. This is because smoothness originated more from an off-road and car tradition and the concept had to be differentiated somehow for all roads and vehicles. For urban traffic, a car/motor vehicle is the worst possible “measuring vehicle”, because everything else on wheels in urban traffic (bicycles, scooters, wheelchairs, strollers, inline skates…) requires a finer differentiation in order for the smoothness concept to offer added value for evaluating road quality.

The result was indeed a partly “stricter” recommendation. The extensive discussions in the course of the smoothness implementation in StreetComplete then had a somewhat corrective effect, so that very_bad, for example, is now a clear exception category in the urban traffic of western cities. While you locate an “anchor point” somewhere between bad and very_bad, I would rather say that the category very_bad itself is an anchor point between “everyday categories” (excellent/good/intermediate/bad) and “off-road categories” (horrible/very_horrible/impassable).

My perception is that the smaller the tire you usually ride on, the higher you set this “anchor point”. With inline skates or a racing bike, the range between excellent and intermediate is enormous, but with a car there are hardly any significant differences perceptible in this range. Accordingly, car drivers tend to map better ratings than cyclists or even skaters. For this reason, we also tried to define further evaluation criteria in the Berlin table, such as vibrations and influence on speed depending on the type of vehicle. Since we have seen in the past that pictures are sometimes not suitable for rating smoothness because they often do not show the actual driving characteristics, I think we should discuss more on such criteria and how to evaluate them.

[rhhs (Richard)]

@Supaplex030

it does fit the existing definition according to vehicle/tire types

Do you think this is true for the very_bad category?

@Supaplex030

This is because smoothness originated more from an off-road and car tradition and the concept had to be differentiated somehow for all roads and vehicles. For urban traffic, a car/motor vehicle is the worst possible “measuring vehicle”, because everything else on wheels in urban traffic (bicycles, scooters, wheelchairs, strollers, inline skates…) requires a finer differentiation in order for the smoothness concept to offer added value for evaluating road quality.

I agree that it’s worth striving for to make smoothness applicable for all roads and vehicles. However in practice this is impossible because the smoothness experience of different vehicles on the same highway is often inevitably different. The difference is biggest for 2-wheeled (single track) vs. 4-wheeled (double track) vehicles:

On this road, cars can hardly avoid the potholes, so for them I would tag this as smoothness=bad, while bicycles, motorbikes, etc. can easily slalom around them so for them I’d tag it as smoothness=good. On the other hand the pavement we discussed here How to map Key:smoothness - #2 by rhhs is quite smooth for cars (the broad tires would hardly notice the lengthwise gaps between the paving stones) while cyclists will have serious problems because a bicycle wheel can fit in those gaps and would make it dangerous to cycle on (similar to how tram rails are dangerous for cyclists).
Maybe we have to make a choice here, and decide that for smoothness below intermediate, the car is the primary “measuring vehicle” and if the smoothness is notably different for 2-wheeled vehicles, we should develop a new tag for it (bike:smoothness=* for instance, where bike = bicycle, motorcycle, etc.?)?

@Supaplex030

The result was indeed a partly “stricter” recommendation. The extensive discussions in the course of the smoothness implementation in StreetComplete then had a somewhat corrective effect, so that very_bad, for example, is now a clear exception category in the urban traffic of western cities.

Would you say that there is now full agreement between the wiki and the Gallery, and that any pictures in the Gallery that are in disagreement with the wiki should be moved/deleted/… ?

@Supaplex030

While you locate an “anchor point” somewhere between bad and very_bad, I would rather say that the category very_bad itself is an anchor point between “everyday categories” (excellent/good/intermediate/bad) and “off-road categories” (horrible/very_horrible/impassable).

By “anchor point” I meant that the category description is as clear as can be and undisputed. For me, very_bad clearly belongs to the “off-road categories” (with very rare exceptions in paved urban environments). I think the concept behind the smoothness classification system is that for every category, there is one or more vehicles for which the category is “usable” and the next worse category is “not usable” (with “usuable” being quite subjective). I think this is what you mean by “anchor point”.

@Supaplex030

For this reason, we also tried to define further evaluation criteria in the Berlin table, such as vibrations and influence on speed depending on the type of vehicle. Since we have seen in the past that pictures are sometimes not suitable for rating smoothness because they often do not show the actual driving characteristics, I think we should discuss more on such criteria and how to evaluate them.

That’s a good idea. I have a key bundle in the glove compartment of my car: if I hardly ever hear it → smoothness=excellent, if I hear it occasionally → smoothness=good, if I hear it all the time → smoothness=intermediate (or worse). Same with the lock chain around the saddle pin of my bicycle. StreetComplete is more visually oriented than I like, also because it’s hard to explain in text in such an app how to evaluate smoothness. Maybe we should develop a tutorial on smoothness that mappers have to pass first before they’re allow to map it

Looking forward to other opinions!

[Matija_Nalis]

@rhhs

I would like to start a discussion on how to resolve the divergence between the two categorisation approaches.

While I agree with some (but not all points) above, and I definitely appreciate the effort into improving smoothness tagging, I’ll note that your approach above appears to be extremely car-centric, only ever mentioning one other mobility type, and that one only in passing.

I’ll suggest to be more inclusive to other traffic on wheels: let’s say 20% cars, 20% motorcycles, 20% bicycles, 20% wheelchairs, and 20% everything else (that categorization is far from ideal, I know, but it would at least promote some wider view). So, for example, if each picture has e.g. 40 words describing that specific smoothness, only about 8 words should be car-related (as a general principle – sometimes it might require little less, sometimes little more, but it can’t be only/mostly about cars for whole tag)

@Supaplex030

For urban traffic, a car/motor vehicle is the worst possible “measuring vehicle”, because everything else on wheels in urban traffic (bicycles, scooters, wheelchairs, strollers, inline skates…) requires a finer differentiation in order for the smoothness concept to offer added value for evaluating road quality.

I agree completely with @Supaplex030 here! As I mentioned in the other thread:

@Matija_Nalis

I think that smoothness should be kept as being useful to the most sensitive vehicle that still can still use that surface, as it means the most difference there.

For example:

• if you drive SUV, you don’t really care very much whether it is good or excellent smoothness of that surface=asphalt. But if you use roller blades, you do care. So the difference between good and excellent should be based primarily on rollerblade users, not SUV users (or other less sensitive vehicles).
• Same for other values, e.g. intermediate vs bad is mostly interesting to about trekking-class bicycle users - it is not interesting for e.g. rollerblade users (as they can’t really use that surface anyway), and it is not really interesting to e.g. tractor users (as it doesn’t make a difference in their lives at all). While SUV is somewhat affected there, that effect is much less then the effect for trekking bicycle user; so in that specific example trekking bicycle user should be the one for whom the difference between that two smoothness values is primarily focused on (as they are about the most sensitive to that specific difference).

[Matija_Nalis]

@rhhs

On this road, cars can hardly avoid the potholes, so for them I would tag this as smoothness=bad, while bicycles, motorbikes, etc. can easily slalom around them so for them I’d tag it as smoothness=good

No, it’s bad really. The fact that someone on rollerblades can easily circle around those does not make it excellent because rollerbladers can use certain parts of it in such manner. Even for bicycles, it’s very taxing, very speed reducing, and quite unlike normal bicycle driving to drive such slalom, so it cannot be good.

Maybe we have to make a choice here, and decide that for smoothness below intermediate, the car is the primary “measuring vehicle” and if the smoothness is notably different for 2-wheeled vehicles, we should develop a new tag for it (bike:smoothness=* for instance, where bike = bicycle, motorcycle, etc.?)?

No, that would be bad idea (as explained in that other thread). For each different smoothness, different vehicle is best measure - for some it is rollerblades, for some road or tracking bicycle, and for some sports car or SUV, etc.

[rhhs (Richard)]

I went back in the history of the table on the smoothness wiki here. I think already then (2008) there was a discrepancy between the text in the “Usable by” column (which hasn’t changed since then) and the photos in the “Photo” column (which have gone through several cycles of replacements). Take the photo for the very_bad category, for instance.

It should be usable by “Car with high clearance, light-duty off road vehicles” but not by “trekking bike, normal cars, Rickshaw”. However I see nothing on the photo that would need high clearance: it can be used by a normal car (slowly) and by a trekking bike.
The other photos have similar problems: the track on the horrible photo can be used by light duty off road vehicles (SUVs), the very_horrible track can be used by a heavy duty off-road vehicle, and the impassable one by a tractor. The bad one can be used by a city bike, and the intermediate one by a racing bike.
(FYI I have experience using normal and SUV cars, and racing and city bikes)

How shall we resolve this issue? Should we stick to the text of the “Usable by” column, and then find photos that illustrate them (this has been my approach so far), or should we re-define “Usable” to something else? “Comfortably usable”, maybe (though that would not apply to normal and high clearance cars, there is no difference in comfort between them on the track shown above)? Or should we increase the number of categories to get the “finer differentiation” that Supaplex said above is needed for evaluating urban road quality? Maybe we could add a category awful for unpaved roads that really need high clearance so very_bad can be used for the worst kind of paved roads (still usable by normal cars and trekking bikes)? I’m not in favour of increasing the number of categories…

[Supaplex030]

@rhhs

or should we re-define “Usable” to something else? “Comfortably usable”, maybe

I’ve always understood “usable” to mean “comfortable usable” (and that’s how we explicitly named it in the Berlin table). Anything else would be strange in my opinion, because with most vehicles you can also use the “worse” smoothness stages, just to a limited extent (e.g. slow or uncomfortable). I think we should formulate this more clearly.

We definitely don’t need new categories because the system basically works well. We just need more/better formulated criteria in order to come to a common understanding of their meaning (especially with different vehicles).

(P. S. To refer to your picture: You can use this road with a normal car, but with very reduced speed and high vibrations. So it’s not (comfortable) usable in a regular manner. Same for a city bike. For me, the picture might show more a bad than a very_bad situation, depending on the amount of loose gravel. So you might need an on-the-ground-experience to categorize it.)

[Hungerburg (Asa Hundert)]

As owner of a trekking bike I’d say, certainly doable that very_bad pictured above, comfortably though? No. Having fun there - sure

Will there ever be a non-subjective measure for smoothness? Perhaps, the key description should start with a description who is target?

[bradrh]

I agree with your suggestion of ‘comfortably useable’. But I’m not sure why you are using the examples the old history.
instead of the wiki page which has better examples?

[rhhs (Richard)]

@Supaplex030

’ve always understood “usable” to mean “comfortable usable” (and that’s how we explicitly named it in the Berlin table). Anything else would be strange in my opinion, because with most vehicles you can also use the “worse” smoothness stages, just to a limited extent (e.g. slow or uncomfortable). I think we should formulate this more clearly.

I now understand that the main reason why I interpreted the wiki text differently is because my interpretation of “usable” is different. I think adding “comfortable” to “usable” makes the tag less objective and verifiable because what is comfortable is likely to be very different for a Papuan trying to reach his village in the jungle on his motorbike, a Russian going mushroom picking with his Lada or a German going shopping in his Mercedes (sorry for the stereotypes…). I think we should strive for that any tag used in OSM means as much as possible the same all over the world, in New Guinea, Russia, Sofia and Berlin.

Interpreting “usable” as “doable” or “passable” makes it more objective and verifiable, at least for cars on unpaved roads. The idea behind my “description” column text is that it adds objective criteria for evaluation, making the tag more verifiable. If you drive an unpaved, quite rough road with a normal car (ground clearance 14 cm) and the underside of the car doesn’t get hit by a stone, rut, etc. then that road is bad or better. If it does get hit, it’s worse than bad. You can then take an SUV with 18 cm clearance and try again: if it passes without a hit, smoothness=very_bad, if there is a hit, take a heavy-duty off-road vehicle and try again, etc.
With the border line between bad and very_bad being anchored like this, I described the better categories according to my own tagging habits as it made sense to me. bad should then be the worst roads that can still be used (though not comfortably) by normal cars, intermediate is intermediate for cars, while good and excellent do not make a difference for cars and can be used to give guidance for skaters, etc. Although I am also a cyclist, I didn’t think my cycling experience could be used. In fact on my racing bike, I’ve ridden many unpaved paths and tracks that I would tag intermediate without wanting to avoid them.

What do you think, does it make sense to define “usable” as “passable” because whether something is passable can be verified (by grond clearance tests, observing actual use, etc.)?

[rhhs (Richard)]

@Hungerburg

Will there ever be a non-subjective measure for smoothness? Perhaps, the key description should start with a description who is target?

No. It is a target worth striving for, but it will never be reached. That doesn’t mean we shouldn’t try.

@bradrh

But I’m not sure why you are using the examples the old history.
instead of the wiki page which has better examples?

Because the newer pictures (and the text in the description column) were mostly added by me, and I wanted to go back to the time before I started to mess up with it

[rhhs (Richard)]

I hope everybody noticed that I put the picture I posted in post 6 up for voting here. It received 17 votes in favour of smoothness=bad and 2 votes for smoothness=very_bad. Can we conclude that because of the very clear majority for bad, most mappers interpret “usable” as “passable/doable” and not as “comfortably usable”?
If so, we need to go through all the pictures in the Gallery and update them where necessary.

[trigpoint]

@rhhs

It received 17 votes in favour of smoothness=bad and 2 votes for smoothness=bad.

Unless I am mistaken they are the same thing.

Although the descriptions are contradictory between bikes and cars.

Whilst a mountain biker (is that the same as trekking biker) would want to route that track for fun, as would a walker/rambler. For normal cars it should only be considered as destination for the last few hundred metres to reach a destination.

[rhhs (Richard)]

Sorry, I edited my mistake: the votes were for bad vs very_bad

I think we should have users who want to reach a destination in mind, not thrill seekers who are looking for bad roads to try their skills.

[julcnx]

@rhhs

However I see nothing on the photo that would need high clearance: it can be used by a normal car (slowly) and by a trekking bike.

I see rocks that could easily result into underbody damage on a standard low-clearance car, hence smoothness=very_bad fits well here.

smoothness=bad says “Good unpaved roads without risk of damage to normal passenger cars are also included here.”

[trigpoint]

And large rocks which won’t do your low profile road tyres much good.

[rhhs (Richard)]

@julcnx

I see rocks that could easily result into underbody damage on a standard low-clearance car, hence smoothness=very_bad fits well here.

A standard car has a clearance of about 14 cm (recent Volkswagen Golf for instance). I see one stone that is a bit bigger than the others, but I don’t think it’s larger than 14 cm.

@trigpoint

And large rocks which won’t do your low profile road tyres much good.

Low profile tyres is something I associate with sports cars (mentioned at smoothness=intermediate)

[julcnx]

@rhhs

A standard car has a clearance of about 14 cm (recent Volkswagen Golf for instance). I see one stone that is a bit bigger than the others, but I don’t think it’s larger than 14 cm.

Have you ever driven on such roads? Rocks will be thrown under body when in contact with moving wheels. If the way is steep, traction may become an issue.

The whole point of the smoothness tag is to present a suitability for most drivers and avoiding damage is an important factor besides high clearance.

Sure this track could be done on a standard car, but it should not be recommended for most drivers.

[mcliquid]

@julcnx

Have you ever driven on such roads?

This is what the majority of Albanian rural roads look like and you don’t drive any special vehicles there. The most frequently seen vehicle in Albania is a Mercedes Benz 124 series with approx. 15cm ground clearance. We were in a Peugeot 308 with 13cm ground clearance. That’s not a problem there and you have to use them to get home.
As on any unsurfaced road, you can hear the stones in the wheel arch, but this does not damage the car. Of course, you reduce your speed to around 30-40 km/h and do not drive the permitted 80 km/h here.

One example here: Mapillary cookie policy use

[rhhs (Richard)]

@julcnx

Have you ever driven on such roads? Rocks will be thrown under body when in contact with moving wheels. If the way is steep, traction may become an issue.

Yes, plenty of times. Loose rocks and lack of traction are not aspects of smoothness, however, they are a property of the surface. I think we should tag smoothness based only on the roughness of the road. A surface=compacted road and a surface=gravel road with equal vertical displacements should be tagged with equal smoothness values. That you will drive slower on a road with loose gravel than on a road where the same size stones are well fixed is due to its surface, not due to its smoothness.

@julcnx

Sure this track could be done on a standard car, but it should not be recommended for most drivers.

That is the issue we are discussing here: should a road be “doable” or “recommended” for a certain vehicle type? I prefer “doable” because it’s a much more verifiable and objective criterion than “recommended” because what someone finds “recommendable” varies much more than what someone finds “doable”.

@mcliquid

This is what the majority of Albanian rural roads look like and you don’t drive any special vehicles there. The most frequently seen vehicle in Albania is a Mercedes Benz 124 series with approx. 15cm ground clearance. We were in a Peugeot 308 with 13cm ground clearance. That’s not a problem there and you have to use them to get home.

I think it would be a useful criterion for determining smoothness: which vehicles do local people actually use on the road? If many people drive a normal car on such a track, then it’s smoothness=bad. If you only meet SUVs and heavy duty off-road vehicles, it’s smoothness=very_bad, etc.