FedHZone: Federated Learning Based Hazardous Zone Detection Using Internet of Drone Things

Air pollution can affect human health, so it is necessary to predict the air quality index (AQI) in advance. In this work, air quality data collected by the Internet of Drone Things (IoDT) is predicted and analyzed to detect highly polluted zones using a distributed privacy-preserving federated learning (FL) approach. Air quality data for a particular zone is collected and then fed into a Long short-term memory (LSTM) based model to predict the pollution level in local training. Global aggregation is done by an IoDT -mounted central server to aggregate the received models from the local IoDT level. Privacy-preserving FedHZone predicts and analyses the AQI values from the past time series dataset in a multi-drone environment. This work detects the most hazardous zone in a larger region of Jaipur, Rajasthan using real-time data. The result shows FedHZone with stacked LSTM having “'2%-7% better accuracy than the vanilla LSTM-based model.