Hi, everybody; I hope this isn’t off topic for this forum.
Motivated by the state of roads in the UK, I’m wondering if anyone is aware of any Open Source/ crowdsourced efforts to assess the condition of surfaces, and then to map them.
I’m aware of lower cost LIDAR equipment, and I believe that some Apple phones have a LIDAR capability.
I’m thinking of something like Mapillary/ Kartaview. Sensor imagery could be gathered, and then scored appropriately, so severity could be seen. I’m thinking that a 100mm pothole on an unclassified and little used road/ lane, would potentially be of less interest/ lower priority than a 50mm pothole on a major motorway/ autobahn/ freeway.
Obviously, potholes are just one example, other immediate possibilities are subsidence, wear and tear, accident damage.
There was a blog post featured in weeklyOSM a while ago about measuring surface smoothness with a smartphone attached to a bicycle, using the vibration sensor.
I’ve experimented with using the iPhone’s LIDAR sensor to measure smoothness of roads/paths. It can produce some interesting results but I’m not sure it’s production-ready. I might open-source the code some time.
Right now i think the accelerometer approach (vibration) is probably more practical.
Re accelerometers, I have some experience of the inverse of this - trying to detect sudden shocks, but making sure that “normal travel” (including along a rough road) didn’t trigger a “shock” alert.
I’m a bit skeptical that there’s much mileage (sorry) in an accelerometer-based approach because people tend to avoid potholes when cycling or driving. As an example, here I’ve added a “surface=bad” tag because of potholes at the edge, but there’s actually a portion in the middle that’s about 1 car wide that people use when there isn’t any oncoming traffic. An accelerometer recording of me cycling down there wouldn’t detect potholes because I’d actually rather stop than go through them.
Regarding ‘mapping potholes’, I expect this to be a layer applied to OSM, not data contained within OSM. It will be open source information, for people that can use it.
I was a co-founder of a start-up that does something similar as Roadroid. Basically we can get accelerometer data from a smartphone to convert to IRI (~road quality index), so road managers can plan maintenance accordingly.
I don’t work there anymore and I don’t know the current state-of-the-art on this matter, but what I do know from experience is that these values are not perfect, but it does work nicely providing an overall of road quality (excellent, good, bad, extremely bad). Some road agencies were (2 years ago) using this kind of technology, on a pilot basis. I am not aware on any road agency using this as a replacement of traditional surveys, nor letting their contractors do that.
On the OSM side, while this excellent/good/bad/extremely bad information can be directly related to the smoothness=* tag, I am not sure if this info can be maintained on a regular basis. For example, on these apps they usually divide surveys into segments (like 20 m/100 m/1 km segments), so one has to group or split (unlikely) that to fit the OSM road segments. Probably it can be done programmatically, trying to create something that matches OSM data, and feed it constantly.
Not sure what you guys think about it, but this is something that cannot be easily done on my point-of-view.
Is the app you worked on open source, need to check Roadroid? If so, it should be possible to compare the features with SmartRoadSense, and see what the next step is.
For example, on these apps their usually divide the surveys into segments (like 20 m/100 m/1 km segments),
ideally they would align to OpenStreetMap way divisions so that we don’t have to split
but it still would be a burden for mappers mapping “manually” because they would not know how to deal with it on way splits
No, Chris, the app I was working on is not open source and given I don’t work there anymore, I can’t do much about that, unfortunately. Regarding features, it was very similar to Roadroid at that time.
Yes, this could be done on the app side, using existing OSM road segments as the “survey segments”. Probably this isn’t super hard to implement, but I’m not sure how open the devs are.
