But, while some companies are retooling the client-side approach, mobile-focused companies are also rethinking location algorithms altogether using machine learning techniques to track indoor location. Some companies think of device location as a complex "DNA chain," whereby using RSSI fingerprints, RSSI trilateration and/or Time difference of arrival (TDoA discussed below) can provide initial location context (i.e. the server tells you where you are); then, by pairing successive RF fingerprints (where is the user walking?) with inertial phone sensors (gyroscope, accelerometer, compass), location can be tracked with very high accuracy, down to 2-3 meter mark.
If that isn't good enough, other mechanisms are added to improve reliability; for example, map processing can also be used to improve accuracy by ruling out impossible paths on the map—also known as error cancellation. But again, one of the limitations to app-based approaches is that not all mobile devices have the same capabilities, so it is more challenging to build an all-inclusive app-based service for all device types (iOS, Android, Windows, etc.).
So the future of Wi-Fi location is clearly focused on mobile devices. But, the trend today is still focused on infrastructure-side location engines. With that in mind, there are a few techniques to discuss.
* Wi-Fi Signal-Based Localization and RF Fingerprinting. RSSI-based localization and RF fingerprinting provide reasonable accuracy, hovering on the disappointing side of room- or aisle-level precision. Without aid from other technologies (exciters, chokepoints, external systems like video systems, advanced antenna systems), 3-10 meter accuracy is about as good as it gets.
With RSSI localization (sometimes called triangulation or trilateration), the key problem is that RF signal strength varies widely at a moment's notice, causing unreliability in measurements. Minimally, three signal sources (APs) are necessary for each measurement, but with varying levels of RF attenuation (due to walls, doors, windows, elevators, etc.) between client and AP, the RSSI-to-distance correlation is somewhat shaky, reducing accuracy.
RF fingerprinting suffers from the same RF variation problems. If you take five "fingerprints" from a single location, the fingerprint will look different each time. Laws of averages help the issue, but it's never perfect. Additionally, RF environments change over weeks and months, so an RF fingerprint taken today may not be valid for that building down the road. Calibration or fingerprinting may become a repetitious process over time.
* Time Difference of Arrival. Time difference of arrival is another technique to determine client location that takes advantage of the constant travelling speed of radio waves, using round-trip time (RTT) of frame exchanges to measure distance. Very fast chip clocks are required to measure nanosecond time granularity; as clock speed increases in Wi-Fi chips in the future, accuracy of TDoA will increase with it.
Sign up for CIO Asia eNewsletters.