How to improve accuracy of indoor positioning?
I should be developing an indoor positioning system for some place , so I started by gathering info about how to develop such a system , the question I am up to now is : what controls the accuracy of positioning and how to improve it ? I found different APIs and projects with various accuracies , for example : ericsson indoor positioning API provides an accuracy within 10 meters , while Qubulus API provides an accuracy within 4 meters , and I met some projects like iDOCNET which claims to provide an accuracy of about 1.2 meters. So, what is the main component of the indoor navigating systems that controls the accuracy ?
Answer
I'm working on a similar project and I tested a couple of the existing tools.
You can find some interesting info about IPS accuracy/precision/resolution here:
Stackoverflow thread on IPS resolution
Read the answer from Carol Politi of TRX Systems, in particular.
In short, the precision depends mainly on the type and density of the radio beacons used as reference signals (that can be GSM/WDCMA/UMTS network cells, wi-fi access points, Bluetooth dongles/beacons, etc.). As a consequence, to improve your system's precision you have to use all of the existing/available radio sources (phone cells AND wi-fi access points) AND, maybe, you have to add/install your own reference points (most likely wi-fi routers).
As long as I was able to see from my experiments so far, the actual precision you can expect from such radio-based systems is as following.
- Phone network: 1 to 2 Km
- Wi-Fi: 10 to 150 m (most often 40 - 150 m)
- Bluetooth: 4 to 10 m (maybe better with Nokia technology, that uses BT 3.0 and special beacons)
Using different technologies together just gives you the precision of the best one. For example, when you use cell phones AND wi-fi access points as reference points, you just get a resolution of 10 to 150 m. Nothing better.
For example/inspiration regarding wi-fi-fingerprint-based systems, look at: Redpin .
The only way to get room-level resolution, using ONLY radio signals (radio multilateration), seems to be Bluetooth. Nokia has developed something for this.
An effective way to improve the resolution of the whole system is to pair a radio-based positioning system (like wi-fi fingerprinting) with a map-based one (Google for "pathfinding": the same map navigation technology used in many 2D games).
This way, you enforce your whole system to pinpoint your user just where he/she can actually be (in a aisle, inside a room), escluding the not-walkable areas (like the inner part of a wall or a not-accessible part of the building). This makes your calculated navigation path much more sensible but in long corridors and aisles the resolution can still be quite bad (5 to 10 m or worse).
Another way is known as "sensor fusion": add to the radio-based system the position/movement knowledge that come from the accelerometer, compass and other sensors that are built-in in the user's mobile device.
Such hybrid systems are already available on the market and can give you a resolution up to 2 - 4 m (room-level, aisle-level) WITHOUT installing any auxiliary radio beacon (such systems are also known as "infrastructureless indoor positioning systems"). A few of these systems use a pathfinding algorithm as well.
For example/inspiration regarding hybrid systems, look at: Footpath .
For an even more inspiring project, see UnLoc by Duke University: UnLoc at Gizmag and UnLoc at Duke .
If you need an even better resolution, most likely you have to install your own Bluetooth beacons (and/or use Nokia technology).