© 2022 , UNSW
COMP9336/4336 Mobile Data Networking
www.cse.unsw.edu.au/~cs9336 or ~cs4336 WiFi Fingerprinting
Adapted from Faria and Cheriton 2006
Copyright By PowCoder代写 加微信 powcoder
Signal fingerprint based positioning
n Received signal is extremely location-specific – dependenceonterrainsandobstacles
n Multipath structure is unique to every location – consideredafingerprintorsignatureofthelocation
n Create fingerprint database for locations of interest n Received signal is matched against database
– toidentifylocationofthetransmittedsignal
© 2022 , UNSW
In most cases, for a given location, RSSI remains within a few dBm of the median value (median shown as ‘0’)
© 2022 , UNSW
L1 could be differentiated from L2 using a single WiFi AP if the RSSI medians were 10dB apart in this case
RSSI Oscillation
Faria and Cheriton 2006
Why a single WiFi AP is not adequate?
n In the previous example, L1 and L2 could not be always separated if the median RSSIs were less than say 5dB
n A single WiFi AP therefore cannot provide high-resolution localization with good accuracy
n What if the mobile device can hear from multiple WiFi APs?
© 2022 , UNSW
Basic WiFi Fingerprinting Example
Fingerprinting Database
RSSI-AP1 RSSI-AP2 RSSI-AP3
Fingerprint (L1) = {-50, -70, -65} Fingerprint (L2) = {-70, -50, -62} Fingerprint (L2) = {-70, -60, -60}
© 2022 , UNSW
Location 1
Location 3
n Vector of RSSIs
n One RSSI for each AP
n Vector length could be variable
Location 2
Median RSSI for AP3 is within 5dB for all three locations, yet the vector of three APs provide unique WiFi fingerprint for these locations!
A basic algorithm for identifying locations with WiFi fingerprint
1. Mobile obtains a real-time fingerprint
2. Compare the real-time fingerprint with each signature in the database (RSSI differences in vector elements)
3. Attach a score to each comparison (number of elements differed less than Δ dBm)
4. Maximum match = signature with max score
© 2022 , UNSW
n 2 signatures in the database for two different locations
– S1={-50,-70,-45}ands2={-40,-70,-35}
n Real-time fingerprint of a mobile = {-44,-66,-34}
n Assuming a Δ=5dBM (needs to be finetuned for real environments)
– ScoreforS1=1,and – ScoreforS2=3
n Maximum match is with location 2 (s2)
n The client positioning is predicted as ‘location 2’
© 2022 , UNSW
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com