Why indoor navigation is so hard

Your phone can get you to the museum, but it can't guide you to the T-Rex.

Map of the Air and Space museum in Washington, D.C. Remember the days before you could pull your smartphone out of your pocket and get instant directions from your current location to anywhere in the world? It’s kind of foggy for me, too.

In fact, I’m so used to relying on my smartphone that I feel increasingly flustered when wandering the aisles of Costco, locating the elephant house at the zoo, or searching for decent food at the airport. Shouldn’t my magical pocket computer help me with this, too?

The answer is “yes,” of course. But there are challenges to implementing indoor navigation today.

User interface

The maps app on your smartphone has one primary concern: getting you from 106 Main Street to 301 Sunny Lane, or from work to home, or from home to Taco Bell. Why are you going to Taco Bell and what percentage of your taco beef will be meat filler? The app doesn’t need to know. Thus, the typical interface for a smartphone maps app is a big map and a search box.

You might assume that an indoor navigation app for, say, the American Museum of Natural History has the same primary concern: getting you from the main entrance to the T-Rex. But why go to the T-Rex? How do I know there’s a T-Rex here anyway? And what if my kids have 20 things they want to see and we only have two hours to see everything? And what’s going on this week — are there special exhibits?

It turns out that creating a useful indoor navigation app requires more than navigation. So, an effective mobile UI should be more “smart guide” and less “paper maps” on your smartphone.

It’s a design challenge, like any other mobile app. Help visitors decide where they need to go first, then direct them there.


Getting directions to the plumbing section of a store is certainly useful. But let’s say you’re looking for a particular Delta kitchen faucet. Wouldn’t it be more useful to search in a retail app for “Delta faucet,” check that it’s in stock, then get directions right to that product? Who cares if it’s in the plumbing section or the kitchen section?

To be truly useful, an app needs to integrate with dynamic data.

Similarly, a university campus app could offer to guide a student to “Kennedy Hall Room 203,” but wouldn’t it be better to search for “Econ 101” instead? Who cares where Econ 101 takes place today? Even better, just have students enter their name once, fetch their schedule, and automatically take them to whatever their next class is. Why make users do more work than they have to?

Current location

OK, so you decide you want directions to that Delta faucet I mentioned earlier. Ideally, the app will automatically start from your current location.

Now comes the great sadness: GPS, as you may know, does not work indoors. The satellite signals are just too weak to penetrate anything much thicker than the metal roof of your car.

However, all modern smartphones have Wi-Fi built in, and wireless networks are common enough in indoor spaces that an app could easily scan for known access points and calculate your position using trilateration.

Here’s the catch, however: Unlike the wide open world of Android, developers on the iPhone side aren’t allowed to perform these Wi-Fi “signal scans.”

Fortunately, there are alternatives. One approach is to make the building do the work instead of the device. Some Wi-Fi installations, such as the Cisco MSE, can determine the location of any wireless device in the building. The access points themselves listen for the Wi-Fi signals created by your phone, then estimate its position via trilateration. This solution has been deployed successfully at a few locations, including at the American Museum of Natural History.

Designing for inaccuracy

One consequence of most indoor positioning systems is a lower degree of accuracy compared to GPS. For instance, indoor systems can usually guess which room you’re in, and that’s about it. Precision depends on signal fluctuations, which depend on factors like how many people are in the room, how you’re holding your phone, and other vagaries.

An effective mobile app must design for this reality from the very beginning. One technique that will help users greatly is to point out quickly recognizable features of the environment.

The Meridian app, for example, uses a short text label to describe each direction step. (Disclosure: I’m the CTO and co-founder of Meridian.) Below, “Rose Room” is clearly marked in the “real world” space and easy to spot, as are the stairs headed down.

Meridian app
The Meridian app uses step-by-step text labels.

The best way to combat inaccuracy, however, is by making it as easy as possible for users to self-correct. In the Meridian app, the map can easily be dragged, rotated, zoomed in and out, and the turn-by-turn steps can be flipped through with ease. If the starting location isn’t perfect, the user will instinctively drag around and figure it out.

Putting it all together

Building amazing indoor app experiences is not only possible, it’s already happening. This year alone, many places — from stadiums and retailers to museums and corporate campuses — have launched apps that are used by hundreds of people every day for navigation and to access location-based content.

Indoor Wi-Fi positioning technology isn’t a research project anymore; it’s out there and works with the devices we all now carry. With the right user interfaces, it can be just as effective as GPS is outdoors.

It’s time to spread the incredible experience of wandering around a place as enormously complex as the History Museum without ever feeling lost.


tags: , , , ,
  • Thanks for writing this, Nick. Great to see more attention on the challenges for Indoor Navigation. Still a lot of missing parts to fill in for this ecosystem to be sustainable (maps, positioning, data integration, services).

    I think scaling to ubiquity for the indoor world presents unique hurdles that were not encountered for the outdoor world. While GPS (and it’s localized cousins) and global outdoor maps certainly weren’t built overnight, they didn’t face the same privacy and accessibility restrictions as indoors. Will users really latch onto indoor navigation if they’re only able to use it in a few places? If not, how will providers scale up their infrastructure to that point?

    Another challenge is meeting user expectations. End users were slowly introduced to the magic of digital outdoor navigation. It wasn’t perfect at first, but it was better than anything they had experienced before. However, now that people are used to the awesome experience of outdoor navigation, how can indoor navigation providers ensure that they don’t disappoint their users with a lackluster initial product?

    It will be exciting to see how the startups and larger companies in the area tackle these problems. Keep up the good work with Meridian!

    -Nick Such
    Co-founder & CEO, BuildingLayer

  • Rich

    If people cannot read signage, why would one think that they are capable of interpreting a tiny (floor) map on a phone? I suppose the phone could speak directions (“U-Turn! U-Turn!”) or maybe squawk or even provide a mild shock if its handler deviates from the correct path to the destination. Phone companies could even charge extra for specialized routing – the ‘express’, ‘scenic’ or ‘back alley’ routes.

    Have we become such mental cripples that the mere navigation through life now requires an electronic guide gadget?

  • The seemingly inexorable march of new technology always brings with it a tinge of melancholy for “simpler times.”

    You’re certainly free to turn off your car’s GPS, shut off your mobile, and wander aimlessly at any time. You may be tempted to switch it back on when you need to find the restroom.

  • Indoor navigation is an exciting step of progress for technology. Hopefully the apps will also be able to detect other people so that those with eyes glued to their smartphone’s map won’t run into the rest of us!

  • Edward

    what about different floor in the same building?
    is Cisco MSE capable of doing trilateration for altitude also?

    I notice AMNH requires you to connect to its wi-fi network first. Is it possible to do indoor navigation on 3G network, with Wi-Fi enabled but not connected to any network?

  • @Edward:

    Yes, the MSE does estimate altitude as well. It actually does its own determination of which floor a device is on and passes that down through the API.

    And you’re correct: Connecting to the wi-fi network is technically not required for the MSE to work. This is the case for some of our other locations.

    AMNH does require you to connect, simply because the MSE server itself is only available for API requests from the local network, and not the public internet.

  • Edward


    thanks for the infomation!

    but I’m still a little bit confused. So you mean the whole story is the MSE “tracks” the iPhone’s location, and then the iPhone client get back the “tracked” location thru the Internet/LAN, but not by the A-GPS / Wifi trilateration from the phone itself?

    Really interesting way to solve the problem!

  • @Edward: you are correct. The device (iPhone or any device with Wi-Fi turned on) is tracked by the network itself. The app simply “asks nicely” for its location from the MSE.

  • Nic

    I think it’s pretty sad that after you mention the hardships of doing such a project with the Iphone, you then say that “there are alternatives…” and the alternative is to have some more complicated work done outside, to get over it! Iphone is _broken_, if they don’t want us to develop that, forget about them.

    The alternative is to move to other more open platforms, maybe Android, but real Linux systems are the best, such as Nokia’s Maemo and N9’s Harmattan. (I’m a “fan”, yeah.)

    But the real reason I am writing is that I noticed you don’t mention Computer Vision. This technology allows for excellent localization, to the accuracy level necessary to create advanced Augmented Reality applications (i.e. excellent localization, including of your device’s orientation). At the same time, it allows the devices to acquire information about the environment, not only map them, but determine states, or recognize objects and track people, etc — more inputs for the dynamic data you mentioned.

    After you look to what Computer Vision can give you, any of these radio-based techniques look like just some extra help… It is great to use it for initial guesses of location, but the real deal will come from the cameras. Heck, there is even a lot we can do for outdoor applications also, but people seem to think GPS+compass+accelerometer is all that is necessary / allowed!

    Join the Computer Vision revolution! It’s all in our hands: the cameras, processors and cloud access.

  • this gps system must be more advanced due to satelight connections

  • Interesting post! Blogged about this topic a long time ago: http://info-architecture.blogspot.com/2010/02/google-indoor-maps.html .
    I think indoor navigation is also very interesting for companies. Most companies have something like a personal directory. The name of the employee, picture, room number, etc. Wouldn’t it be create if you could get navigation help to go to the person you have a meeting with? In large companies this is an issue.
    Even more interesting, I think, would be to add an expertise layer to those maps. Then you could say: I’m looking for the closest person with expertise x. Then a map is shown with that person on it and navigation info how to get to him/her.
    Curious to hear what you all think of these ideas!

  • Would the creation of a new Bluetooth Location profile aid with this? Simple, low cost (and low power) location ids could be put in place around a facility (e.g., “doorframe” or “floorpad” devices), pushing a fixed, known location (and possibly the URL of a reference map). The phone would receive this and display the results accordingly.

    Problems here include adoption of YABP (Yet Another Bluetooth Profile™) by phone manufacturers, as well as how to allow facility owners to easily map the fixed locations onto the “doorframe/floorpad” devices. Also, does Bluetooth allow for some form of quick-pairing/release (and triangulation if multiple signals are detected)?

  • @Samuel: That is a fantastic idea! We’re actually working with a corporate location to add some of these exact features, conference-room booking to name one.

    @John: I think Apple agrees with you on that actually. The new iPhone 4S supports the Bluetooth 4.0 spec which allows for low-power “beacons” to be placed and scanned for location-determination. We’re really excited for this to become a standard feature among smartphones.

  • Adi

    Very interesting post, Nick.

    I was wondering, is it possible to somehow simulate a GPS protocol inside a building (instead of using WIFI)?

  • @Adi: I believe it should be technically possible, yes.

    Over the years, we’ve looked for anyone building GPS repeaters but we never found anything off-the-shelf that we could buy and install.

    I do believe the government heavily restricts that portion of the spectrum since GPS exists primarily for the military. So this might be a reason.

  • HP’s Cambridge Research Lab tried triangulation from wifi. Turns out to work very poorly in a space with corridors. They tend to channel the wifi signal. You end up needing to create a set of tuples per location, calibrating the entire space. And of course, when you move big things around, it changes.

  • Good article. I think most people prob have WiFi settings to not notify them on new WiFi hotspots all the time, since that gets kind of annoying. So betting on WiFi is not a winner in many cases. There was that company in the UK that would do that via cell towers that were installed in buildings. That would universally. Unfortunately I forgot the name.

  • The innate inaccuracy of WiFi-based positioning is particularly apparent when talking about elevation, since a few feet of error can easily put you a floor or two off in terms of your position in a building. It’s better than nothing, but not suitable for anything that really matters.

    There are a number of speciality solutions for higher resolution indoor location tracking, however. WhereNet, Metratec, Agilion, and others offer solutions in that space.

  • Gagan Rajpal

    For User Interface, the way to do indoor navigation is through voice (e.g. Siri) not visual. Visual maps are distracting while navigating indoors. In-fact visual maps a are distracting anytime you are trying to find something while you are driving, riding, or walking ..
    When I am driving I prefer voice as the primary for of output from my GPS with the map on it as a hint.

    A voice interface would be the perfect way for this. It can much more easily allow the user to provide more information about the environment to the navigation system.

  • Great post, I too was curious about this technology and wrote about a few other approaches on ZDNet:


  • Hey Gagan, tell us what you think.

  • Bit late, but this announcement has some relevance here: http://www.opengeospatial.org/pressroom/pressreleases/1369

    You can download the discussion paper from the OGC website or using the link in the PR.

  • Paul

    Hi and thanks for the neat summary of the issues. I have been researching ways to build indoor navigation apps for a while now. I am not a coder but its just a hobby. I am right now finalising an indoor navigation app of a very large US multi-level mall. The app gives turn-by-turn directions plus distances along the shortest route. On the next rebuild I am going to add voice-aided navigation. Its been a challenge but I have enjoyed every minute of it. OK theres no automatic location detection as I dont have the technology know-how and wi-fi indoors isnt very prevalent in my country so testing wi fi location here is a problem. The app will run on the i-phone, Android and Symbain 60 devices. I hope to soon extend this to Windows Phone 7 devices. I am glad to be a part of the indoor navigation revolution.

  • Paul

    OK so heres an issue that seems to negate the whole business. Yes its very difficult to get indoor navigation correct. Strictly speaking until we can locate our position automatically its not true navigation but simply mapping. Now when you get it right and use wifie or whatever to locate your self and you think you have TRUE navigation then what happens ? You get sued by people who are upset that you know where their phone is inside the building. Then there are companies ranting and raving about being able to locate traffic through a mall and provide stats on these flows and to be able to tell stores that Mr Jones is in aisle three and he likes fishing. I think this is a BRILLIANT idea and would love to see the stats. But what happens – people are again upset that you can find their phone indoors and so they switch it off in the mall ! Guess we just cant win can we ??? Still I’m not giving up !

  • Rossy

    I was wondering what method to use if you want to locate someone else in a room or building and showing you the distance that person is away from you.

    Could someone help me out with this one?