Cattle tracking with LoRa and machine learning
To curb the increase in livestock theft in Sub-saharan countries, this project looked at machine learning applications like cattle tracking on a farm. Received signal strength indicators from home available WiFis were utilized to fingerprint locations on the farm. Where WiFis were not available, ESP866 Node MCUs were used to deploy access points. These ESP8266 Node MCUs were also given SSIDs that could easily be used for classification using the SSIDs as labels in the program. Two classifiers were explored, the support vector machine (SVM) and the Decision Tree. A linear kernel with a very low gamma value of 0.001 was used for the SVM classifier. The decision tree classifier yielded an average accuracy of 0.89683659, while the support vector machine yielded an average accuracy of 0.915699. The SVM classifier emerged as a suitable classifier to use to achieve reliable accuracy in cattle tracking. Lora proved to be an essential tool in wireless communication of the prediction results from the pendant/transmitter designed to the receiver, hence enhancing mobility in tracking. Future works include incorporating distance calculation to make the tracking process more straightforward and efficient. Sensor nodes that will further provide the machine learning algorithm designed with unique data values will also be incorporated to compensate for longer farms that might need the deployment of more Node MCU access points. This is an effort to maintain low costs in tracking using the solution provided by this project. Therefore, the solution presented by this project proved to be feasible.
Capstone Project submitted to the Department of Engineering, Ashesi University in partial fulfillment of the requirements for the award of Bachelor of Science degree in Computer Engineering, May 2021
location tracking, livestock tracking