An Adaptive Drone-based Multigenerational Sensor System for Monitoring Large Ecosystems

With the increase in the use of sensors and drones to monitor soil and crop conditions in vast agricultural lands as well as for situational-monitoring of natural disasters such as forest fires and increasing sea levels, there is a need for systems that can support long-term monitoring with minimal cost. In this paper, we propose a novel architecture that combines sensor, drone, blockchain, and machine learning technologies to support a long-term monitoring system for vast areas of land. Long-term monitoring is achieved by employing multigenerational sensors that have been deployed at the start but are woken up in stages as and when needed to replace or reinforce existing sensors. A swarm of drones periodically scans the landscape to identify areas that need immediate attention (e.g., dry soil, pests or insects, humidity, temperature, etc.). Drones store and process the data using onboard units and make real-time decisions regarding changing path plans or changing scanning frequencies. Optionally, data is sent to the base station that fuses the information from multiple drones in the swarm, and inputs the fused data to a machine-learning model that recommends the needed actions. The actions may include obtaining additional data, awakening more sensors, actions to handle any situation by signaling for external actions such as cloud seeding, controlled chemical spraying from a drone, etc., in agricultural applications. Using simulated scenarios, we have employed three prediction algorithms—the Prophet model, CNN, and LSTM—to determine the effectiveness of these algorithms in predicting sensors values in the simulated scenarios. The proposed architecture is scalable, supports long-term monitoring, and is effective in predicting future conditions to enable preemptive actions.