Header image: SLAM Crowd Engine crowd-sourcing recording data from active smartphones.

Indoor Location and Positioning services have taken on an important role for building management for technology driven companies. Ever since Indoor Location was in its baby shoes, several different approaches and methods have evolved in search of the best and most efficient technological solution. One of the most common methods is based on RSSI (Received Signal Strength Indicator) and this popularity is because of there being no need for special hardware.

 

Radio maps, which are indispensable for enabling RSSI based localization services, are normally created by taking measurements of every position inside a building. As this is often connected to a heavy manual workload, thus highly time consuming and costly, researchers are continuously trying to find more efficient ways to take these measurements. In order to address this problem, indoo.rs has developed a multi-stage SLAM (Simultaneous Localization and Mapping) technology. The first two stages shall be briefly explained.

 

Whereas the SLAM Engine automates and accelerates the majority of the radio mapping process through dedicated recordings, the SLAM Crowd Engine (SCE) steadily improves the map and keeps the map up to date through crowd-sourcing data. More precisely, the SCE uses crowd-sourced recordings from users navigating inside the building (see header image).

 

However, since there’s usually a broad range of different smartphone types in use, it is unlikely that all of the collected data share the same RSSI characteristics. This is where indoo.rs CaLibre™ comes into play.

In general, indoo.rs CaLibre™ calibrates the RSSI of a crowd-sourced recording on the fly, only through the data within the recording. It doesn’t require any pre-knowledge of the device in use nor sharing any calibration results between recordings. Therefore, it’s a “memory-free” solution that spares the need of updates or maintenance of a database of calibrated devices.

 

To achieve “memory-free” calibration, the CaLibre™ manages to use the radio information from recordings more efficiently by “scan tiling”. In the scan tiling process, the recordings are grouped together into small tiles according to their visible network. All scans within a tile will be rather close to each other and thus have a similar power level. This is shown in the figure below.

 

indoo.rs CaLibre tile matching
Figure: Matching tiles in the Calibre algorithm

The RSSI characteristics from different recordings will then be scaled, so that their RSSI readings will match each other for each tile (this is due to the assumption, that the power level within a tile is consistent) In the last step of the CaLibre™, the scaling factor will be applied to rescale the corresponding recording, which makes sure that the RSSI readings indicate the same power level as the standard reference.    

 

CaLibre™ Benefits in a nutshell:

  • CaLibre™ calibrates each crowd recording on the fly and thus doesn’t require any database for calibration results. (Construction and maintenance costs of such a database are usually extremely high.)
  • CaLibre™ calibrates any incoming recording with no pre-knowledge needed, which guarantees that each recording can be calibrated and no information will be wasted.
  • The fact that each recording is treated independently guarantees that a potential calibration error will not be spread across the boundary of the recording and makes CaLibre™ more robust to the calibration inaccuracy.

 

If you are interested in CaLibre™ please download the paper written by
Boxian Dong, Thomas Burgess, and Hans-Berndt Neuner