How to zoom in to centimeter-level scale in OpenStreetMap?

I am using OpenStreetMap and it’s API to generate routes and waypoints for my robot. However I cannot zoom in the map more than the scale of 20m?
I am using RTK-GPS to accurately measure my position to centimeter-scale and want to see and use coordinates at that level. Is this possible?
If not, is there any other open-source or even proprietary service that is good for this use-case?

P.s.: Another problem I am facing is that when I plug coordinates from my sensor to the map, it has an offset and does not point to the location where my robot is standing. Is this due to bad mapping? Any way i can resolve this? Any service where I wouldn’t face this problem. The offset is of the order of meters

Typically OSM based raster maps provide tiles up to zoom levels between 18 and 22, editors will allow you to overzoom more, but most web maps don’t do that. I would suggest setting up your own private server using vector tiles as a simple solution around that, see Vector tiles - OpenStreetMap Wiki


There is no one number for accuracy of OpenStreetMap data

  • OpenStreetMap coordinates have limited precision ~1cm at the equator less at higher latitudes.
  • OpenStreetMap stores coordinates in WGS84, not in an continent anchored ITRF system, while this mostly only has small effects, there are places, for example Australia, where this will lead to mapped OSM elements moving noticeably relative to the coordinate system over time.
  • the accuracy with which objects are mapped depends strongly on how the data was collected/produced. This can be a couple of dozen of meters for GNSS in a dense forest or urban canyon, over perhaps 2-3 meters using very high quality aerial imagery (even the best sources will quote around 1-2m in hilly areas), down to sub 1 meter using specialized surveying equipment.
  • and there are numerous other things that can impact the accuracy on top of all of this.

With other words you will have to determine what applies to your area yourself (and potentially improve the mapping there).

  1. Technically there are vector tiles provider, including a Carto style by none other than Esri. But OSM data isn’t what you want, as detailed.
  2. Your RTK already gives you the coords. Any map is for visualization or reference. Looking at aerial imagery is another possibility. That’s better than having only OSM. A hybrid view with vector map overlay is the best how you can use OSM to help.
  3. For navigation on solely drawn maps, your best best is government maps in survey sheet and engineering plan quality. That’s essentially what you are doing anyway. If you tell us what country you are in, others may be able to help. Alternatively if government open data is not highly available, when you are only working in a certain location, you might ask the site eg the university to give you drawings to georeference them, I guess (not working in this field). Can you tell us how large an area you are working in?
  4. Unclear what offset you mean. I doubt you can tell your position in metres by looking at OSM in the first place. In general, you should start at a position of known coords viz survey markers. Is your RTK floating or fixed, or actually only at DGPS level? As mentioned, specifically Australia has plate drifts requiring the use of a plate-fixed datum. RTK will still suffer from urban canyon multipath between buildings especially L2 frequency, although it may not be as severe as you think particularly between mid-rise. Can you afford L5 if either your receiver or antenna isn’t already tri-band, or you are already using L1 + L5 dual-band? Do you have ADR (wheel ticks) or UDR (IMU) dead reckoning on your robot?

I have tried using RTK for editing: Don’t use OSM for that.

You certainly can create maps that show OpenStreetMap data at higher zoom levels than any of the layers at support (21) - this is an example at zoom 24. A rough bit of mental arithmetic suggests that 1 screen cm is about 25cm on the ground. Those are actually raster maps, but as Simon says above a vector approach makes more sense at high zoom levels because an increase in client zoom can be easily shown without extra data.

With regard to routing, OSM doesn’t have an API, it just passes data to one of three external routers (see “Directions courtesy of…” at the bottom of a route list displayed in OSM). I would expect that in addition to your own map tiles you’re going to need to fix some way to pass co-ordinates to a router.

Thank you for your quick responses, and your explanations of the working of OSM. I will look into vector tiles and aerial imagery available.
We are working in a university in India. It turns out that dense tree canopy over certain roads already blocks out gps signals and rtk won’t help to increase accuracy there. Our goal was to accomplish localization of our autonomous robot using gps, and thus needed a highly accurate map, which does not seem possible and practical to pursue long term, thus, we are looking at other strategies.
Thank you for your help again, we will surely try to improve the OSM maps in our area!