Robust and Undemanding WiFi-fingerprint based Indoor Localization with Independent Access Points

Our proposed localization system is similar to existing WiFi-fingerprint localization systems. It works by first collecting received signal strength (RSS) values of access points (APs) from different reference points in the area of interest. This initial training can be done either manually or by crowd-sourcing. The WiFi-fingerprint map is generated from collected RSSs at all reference points. However, instead of utilizing unreliable absolute RSS values, our system uses RSS differences between every pair of APs as a fingerprint metric. This fingerprint metric is denoted as AP-Sequence. The AP-Sequence is used to partition the area of interest into small regions. Each small region is associated with unique ordered RSS sequence. The WiFi-fingerprint map consists of this small regions. Figure 1 illustrates a high-level application scenario of APSequence fingerprint localization. When a user A wants to know her location, she first scans the RSS values of APs in her proximity. For example, the user A scans wireless channels and observes AP1, AP2 and AP3 with RSS values of −42 dBm, −65 dBm, and −72 dBm, respectively. Based on the relative difference between RSS values of the three APs, the AP-Sequence of <1, 2, 3> is generated. Essentially, it is an ordered sequence of APs from high to low in terms of RSS strength. Then the user A uploads AP-Sequence <1, 2, 3> to AP-Sequence fingerprint map server for localization. The AP-Sequence of <1, 2, 3> is compared to AP-Sequence fingerprint map and she is ultimately localized to a region associated with the best matching AP-Sequence.