Help with BlindAid: predicting future 3D data formats & general advice

I am a Research Scientist at MIT working on a new assistive device for the blind called BlindAid. Several people have suggested that OSM may be the best option for getting map data to be read by our system. Unfortunately, OSM currently appears to fall short of our needs, so I must try to predict what it–or some other source of map data–will be like in the future. I would be grateful for any advice or help with this task. If you know people who might be able to help, please alert them. Some OSM groups that appear to have overlapping interests are:

  • 3D Development
  • OSM for the blind
  • Indoor Mapping (OSM-ManSizedMatters project)
  • WikiProject_Metro_systems
  • Accessibility

Following this introductory note is some project background and then specific questions. Please feel free to respond to any or all of the questions and forgive me for the long list, or if I am posting this in the wrong place! I am new to OSM. I image that I will have more questions in the future and I would welcome guidance on how best to interact with the OSM community.

You may contact me directly at schloerbATmitDOTedu if you prefer.

Thanks,
David

BACKGROUND

We often describe our system as a kind of MapQuest or Google Maps for the blind, that blind people may use to learn about new environments (e.g., city streets, libraries, bus stations) on their own before actually visiting the places. The system employs a desktop robotic device–a “haptic” device–that functions like a miniature white cane and allows the user to explore virtual maps of the world by touch. The specific haptic device we use is the Phantom (Sensable Technologies) and we use OpenHaptics, which has calls similar to OpenGL, to render the haptic scene. The user also hears spatialized sounds as if standing in the virtual world. I would be happy to send you a copy of a paper that describes our initial prototype in more detail.

We recently got funding to continue the project and my main goal at the moment is to find a source of comprehensive digital maps that may be read by the system. Currently, we create our virtual maps manually, which severely limits the potential application. Indeed, our system will never be truly useful without a good source of maps.

Ideally, I would like to find “maps” that actually define 3D virtual environments with sounds, models of buildings, interior building layouts, and with appropriate scale/detail to meet the navigational needs of a person finding his/her way with a white cane or guide dog (walking, via mass transit, or riding in a car). Unfortunately, none of the sources of maps that I have investigated so far appear suitable.

My strategy is to assume that someone (OSM or perhaps a commercial source) will eventually provide the detailed 3D maps we need. My task is to try to predict the data format(s) that will be used so that we can do our development in parallel. Presumably, any data format selected now will be revised in the future, but if we pick a good one that becomes standard it will be much easier to adapt to a future version than if our initial choice becomes obsolete. And, even if OSM (or other source) provides its map data in a completely different format in the future, it should be possible to find software that will convert the available data into the format we select now if we pick a good/standard one.

SPECIFIC QUESTIONS

  1. What 3D map data format(s) should we choose for our application and why? Do conversion utilities exist now for transforming the format you propose to/from other standard types?

  2. Is my strategy reasonable? Perhaps a 3D map database with sufficient detail for people who are blind will never be developed. Perhaps it just isn’t possible to predict which data format(s) will become standard. Can you suggest another approach?

  3. Is OSM our best hope for getting the needed 3D map data in the future or is there a better source? Indeed, does a suitable source exist now and, if so, what is it?

  4. Is it possible to access commercial map databases like MapQuest or Google Earth at a low enough level to render a map on my local computer (e.g., using OpenGL)? Even if we had to pay for the service, it might be our best long-term option. Unfortunately, my understanding is that they only provide rasterized images.

  5. Is there any kind of plan or consensus within OSM toward developing detailed 3D maps? More generally, what is the vision for the future? What data format(s) are being proposed?

  6. I have been told that there is a great deal of usable navigation/accessibility data on the web, but to find it you need to go to specific sites hosted by many different groups. For example, http://easyaccesschicago.org/ provides a wealth of accessibility information in the Chicago area. An argument for selecting OSM is that it may become a repository for combining such data in a single searchable location. Is this being done?

  7. I have read that XML based formats (e.g., GML, KML, CityGML) may not be adequate for detailed geometry because of the large files that are needed. Is this true? Is there is a way to access only the data that is actually needed at a given moment or a way to select/parse-out only data that is relevant to the blind?

  8. Perhaps it will be necessary to read multiple data formats from different sources. For example, the OSM-3D project gets elevation data from NASA (STRM). Is there any kind of hybrid database standard that incorporates a collection of different types of formats? I could imagine, for example, an OSM map giving a course outline of streets and buildings that links to a detailed CAD model when the user wants to explore a specific building.

  9. Another approach might be for us to convert our system to read 2D maps. It would have 3D structure like a tactile map, but we would not attempt to create a complete 3D world. Does anyone have arguments for or against this approach, particularly in terms of what types of map data exist now or might reasonably be expected to be available in the future?

  10. BlindAid allows the user to listen to sounds as if standing in the virtual world. Unfortunately, while pictures are often incorporated into electronic maps, I haven’t found anyone who includes sounds. It seems like it would be fairly easy to record sounds on a given street or near points of interest that users could listen to when they explore the map. Is anybody thinking about this or doing it?

Wow what a lot of questions: most of which look like the really justify in-depth answers. Quickly, my 2 cents.

At the moment I think OSM is pretty much the only game in town for the source of much of the regular map data which you want. That doesn’t mean it can do everything, but most other sources will either be encumbered by copyright, legal, technical, or financial restrictions. There is already a technical eco-system for mapping for the blind and wheelchair users. Probably the email list devoted to the Loro-Dux project (http://wiki.openstreetmap.org/wiki/LoroDux) might be a better place to get detailed answers.

It’s impossible to cover the range of things done by OSM or its possibilities in a message. The best way to find out these things might be to attend the State of the Map conference in Denver this September: http://wiki.openstreetmap.org/wiki/State_Of_The_Map_2011. Last year there was a significant presence from people interested in similar problems.

You will almost certainly need your own format for data on the device: the main OSM XML format is designed for moving data from place to place. Most applications have separate internal formats.

A lot of data should probably be held externally to the map data (sounds, buildings etc): a problem arises of ensuring that the linkage between the map data and external files remains persistent. OSM does not provide a mechanism to do this at present, although I believe there is a ground swell of interest in working out how this might be done.

The OpenSource flight simulator groups (X-Plane and FlightGear) are either using or planning to use OSM data in 3-D apps: there may be some common technical ground.

I’ll try to add a few answers to your questions, too.

Generally, many of your questions are related to data formats. I believe that the choice of data format used in most parts of your application is independent from the decision whether or not you use OpenStreetMap data. If you use OSM, you will read OSM data, but you will probably not use OSM’s data formats elsewhere else in your application or service. OSM data comes in a format optimized primarily for editing and exchange between the different applications in the OSM ecosystem. My suggestion is to choose a format dedicated to three-dimensional representation instead and convert OSM data to that format.

As there is no single entity controlling the direction of the OpenStreetMap project, this mostly depends on the personal goals of the mappers. It seems that many mappers are in fact interested in adding detail usable for 3D visualization to OpenStreetMap, especially in those parts of the world where OSM has already achieved a good degree of completeness for basic featues.

This means that within a few years, you might find

  • almost-complete coverage of building outlines thanks to aerial imagery tracing
  • heights and entrance positions for the more interesting buildings and structures, occasionally per-level outlines or building parts
  • a small number indoor maps in places such as universities or malls
  • spotty coverage with outlines for irregularly shaped linear features (such as some roads)
    in well-mapped areas. You will probably not get absolute elevation, with a few exceptions such as mountain peaks.

That data will likely be made to fit into the current OSM data format, but the tagging is still not clear for some of these features. Most of the relevant conventions are still only suggestions.

Services that use OSM often try to give data, bug reports or other improvements back to OSM. For example, http://wheelmap.org lets users add basic wheelchair-related accessibility information using a tool that is ultimately just a very simple OpenStreetMap editing frontend.

If you want absolute terrain elevation or building shapes beyond the extruded polygons that can be derived from OSM data, then yes, you will need additional data sources.

I’m currently experimenting with conversion from OSM data to 3D models (see osm2world.org for my first clumsy results). I think that OSM data is very promising for the task to get small-scale details right (e.g. make sure that steps go up the right way, that tunnels and bridges are represented correctly etc.), but I’ve come to the conclusion that I will need external data sources like SRTM to improve the terrain model.

I just want to thank SK53 and Tordanik for their helpful responses. I’ve waited until now to say this because I was hoping to hear from some other people. In any event, I am planning to take SK53’s advice about attending the State of the Map conference in Denver. I look forward to learning more about OSM then!

Great stuff guys! …But;

  1. It leaves my question about the existence of a “width” key unanswered. Does a legitimate, built in, or generally accepted key describing “Street width” or just plain “width” …exist?

  2. I am no expert or spokes-person for X-plane, but yes X-Plane certainly makes use of open street map for roads and probably land classes…

  3. It’s good to hear there is more recognition for the need of coordinating hi-Res height data Visa vis, the SRTM Data from NASA… which has been available via the USGS for close to 30 years! I find it ludicrous to see the power and breadth and possibility of OSM, and yet the Org, has no official merging of vector and height dat in that time??

SRTM data continues to be improved, and has always, since The space shuttle went up with two powerful, high-resolution Radar Sets to terrain map 90% of the surface of the earth in 1990.

Chas

David, I think you would be well served to check out ArcGIS… they deal with commercial outfits using satellite data meaning satellite photographs and Gps Geo location with a much higher granularity and I am pretty sure that they would be the appropriate source for creating the kind of data that you need…much better than open street map can or has available…

Mind you… looking at OSM’s date-ref format, XML, there is no reason why the height data that you’re looking for could not be included in OSM’s as per my comment regarding the same posted this date here.

As always, in any form in which I participate, I am happy to be shown the error of my thinking…

Good luck with your projects with your sight-challenged folks.

Regards,

Chas

Why did you hijack an old thread to re-ask your question ?
You can use the width tag to map the width of a street.

@schloerb you don’t actually specify what level of detail / feature size / positional accuracy you are aiming for. that makes it a bit difficult to determine if any kind of generalized representation can actually cover your needs, or if you should be looking at direct representations (3d point clouds or similar for example) of the objects.

I have counted the steps of every public stair in the inner ring of Brussels.
Every public stair has some spheric pictures tagged to it.

Not sure whether you are still working on this, but ‘StreetComplete’ (Android) may be able to help. See my post here:
https://forum.openstreetmap.org/viewtopic.php?pid=652902#p652902

Thanks,

Chris
chris_debian
2E0FRU