Development of low-cost Lora rssi based detection of cattle rustling: a case study in Laos

Abstract

This thesis investigates detection of cattle rustling by using low-cost Internet-of-Things (IoT) equipment. Recently, cattle localization based on LoRa received signal strength indicator (RSSI) values has gained a lot of attention since it requires low-power consumption and supports long-range communications. However, LoRa RSSI based localization is known to be inaccurate due to the frequency-changing nature of LoRa signals. Therefore, instead of trying to estimate an exact location, this paper focuses on detection whether an animal-attached sensor node is inside or outside the fence area. We propose the use of transmitter reference points (RPs) installed in selected locations to help determine the status of each animal. Experimental results indicate that, when received RSSIs from the same transmitter-receiver locations change over days, the use of RSSIs from transmitter RPs instead of RSSIs from a previous day as the train dataset allows the detection algorithm to adapt over time and provide higher detection accuracies in comparison to detection without using transmitter RPs. Finally, comparisons among well-known classification algorithms, including k-nearest neighbors (k-NN), support vector machine (SVM), decision tree, and random forest, are performed in terms of detection accuracies.