Prediction of WiFi Signal using Kalman Filter for Fingerprinting-based Mobile Robot Wireless Positioning System

Abstract

The non-predictive behaviour of wireless signal especially that of 2.4GHz WiFi due to the complex signal propagation is primarily non-usable for mobile robot positioning system. It is fluctuating and prone to error that made positioning accuracy haywires even in stationary location. Therefore, there is a need to estimate the wireless signal to its real value as per fingerprint location. This paper proposed to use the Linear Kalman Filter (LKF) to predict the wireless signal, i.e. the WiFi Received Signal Strength (RSS) to estimate the location using the Weighted K-Nearest Neighbor (WKNN) algorithm that matched the fingerprinting database constructed beforehand. By employing the LKF, the accuracy of the positioning system at any stationary location has improved significantly when compared to the use of raw original WiFi signal.