Impact of APs selection scheme on compressive sensing-fingerprinting based IPS performance

Abstract

Improving the accuracy of indoor positioning systems (IPS) has constantly been and still an active challenge. The Compressive Sensing (CS) approach has recently been shown to have a potential for indoor positioning systems (IPSs) solution once the positioning problem is modeled into a sparse signal recovery one. Indeed, positioning schemes using Received Signal Strength Indicator (RSSI) fingerprint maps and CS methods show remarkable improvements in performance. In this paper, we are presenting the impact of the Access Points (APs) selection scheme on the performance of both CS and k-Nearest Neighbor (k-NN) RSSI-based IPSs methods. The performance analysis has been investigated using three different APs selection schemes, namely, the Probability of Detection (POD), the Time Variance, and the Random selection schemes. Our simulation results show that the Random APs selection scheme outperforms POD and the Time Variance schemes when CS-based framework is used while k-NN outperforms CS with POD and Time Variance schemes. Results also show that CS outperforms the k-NN method under the same number of selected APs when using the Random APs selection.