AHSWN Home • Issue Contents • Forthcoming Papers
RSSI-based Localization using K-Nearest Neighbors
Many popular location-based applications use Received-Signal-Strength-Indicator (RSSI) measurement to obtain location because it does not require special hardware. However, physical phenomena affect the propagation of the radio signal, which causes noise in RSSI measurements. In this paper, to account for signal noise due to propagation, RSSI measurements are represented by distance intervals. Then, based on the defined distance intervals, the K-Nearest Neighbors (KNN) machine learning technique is used to improve the accuracy of RSSI-based localization methods. For performance evaluation, several testbeds were used under different conditions, such as outdoor/indoor environment (Residence, Laboratory, Library, etc.), type of equipment (Smartphones, Tmotesky, Telosb, etc.), and communication technologies (Wifi, Zigbee, BLE, etc.). The experimental results show that the proposed method significantly reduces the localization error in these aspects compared to other well-known RSSI-based localization methods.
Keywords: Received Signal Strength Indicator (RSSI), Distance Interval, K-Nearest Neighbors, Wireless Networks, Disk-based Multilateration, RSSI-based Range based Localization, Indoor/Outdoor Localization.