Improving LoRaWAN Fingerprint-Based Localization by Detecting and Eliminating Noisy RSSI Measurements

Industry 5.0 (I5.0), is gaining more attention as a means of improving human life in the future. IoT plays a critical role in providing the necessary information and performing actions. Based on the wide range of industries, LPWAN and more specifically LoRaWAN is considered as a great progress to activate the IoT in the industry. However, to make the information provided by IoT meaningful, a localization method for end nodes is necessary. Fingerprint-based mapping is one of the developed methods for localization. Nonetheless, the frequent updates required for the fingerprint map pose a significant challenge. Additionally, the presence of noise in the fingerprint map, caused by environmental factors, significantly impacts localization accuracy. So, it is important to learn the environmental effects on the Received Signal Strength Indicator (RSSI) for performing a real differentiation between noisy and noiseless RSSI during the fingerprint map construction stage. By constructing an environment-aware fingerprint map and emphasizing the importance of noisy RSSI detection and elimination, this research contributes to enhancing the reliability and precision of LoRaWAN-based fingerprint-based localization systems by leveraging image processing techniques. For the evaluation step, a real implementation of proposed method in a car parking environment serves as testbed, where the influence of vehicular movements on RSSI measurements is prominent. The experimental results demonstrate that 83 percent accuracy in detecting noisy RSSI can be achieved by using the SVM classifier. This distinction between noisy and noiseless RSSI contributes to a significant improvement of 36 percent in location estimation accuracy.