Optimizing Data Collection in IoT Networks with LoRa Equipped Drone: Minimizing Age of Information through Reinforcement Learning

This thesis considers a UAV-assisted IoT network, where an unmanned aerial vehicle (UAV) is deployed to collect status update data from various heterogeneous sensors monitoring physical processes. The UAV then transmits the collected data to a base station using LoRa (Long Range) communication, which is known for its low-power, low-cost, and long-range capabilities. However, in urban environments, the effectiveness of LoRa can be significantly compromised due to diverse physical settings, signal interference, and obstructions, making it difficult to determine the optimal locations for transferring the collected data to the base station. The primary objective of this thesis is to develop a comprehensive system that leverages a UAV to optimize data collection from sensor networks within an unknown environment, where the LoRa signal qualities of different locations are not predetermined. Specifically, this work aims to minimize the Age of Information (AoI) to ensure that the data received at the base station is as fresh as possible, thereby enhancing the timeliness and relevance of the information. A critical challenge in achieving this objective is the joint optimization of the UAV's trajectory and the selection of data transmission locations to minimize the AoI. We formulate this as a new optimization problem and demonstrate that it is NP-Hard. First, to understand how the real environment affects the LoRa connection, a testbed with a LoRa transmitter and receiver to collect real-world transmission data in a 1 km² urban area has been developed. Using this collected data, an initial solution was implemented with Gurobi, based on the assumption of a perfectly known data rate distribution. Next, a a real-world scenario where the LoRa signal quality is unknown at different locations has been considered. Given the high computational cost of the Gurobi solution, we propose a heuristic algorithm that leverages reinforcement learning to learn the environment and plan the UAV trajectory with the objective of minimizing the AoI. Additionally, we performed extensive experiments to assess the performance of the proposed solution against existing approaches. In conclusion, this thesis develops an efficient UAV-aided data collection system with LoRa communication capabilities for base stations in urban environments. By jointly optimizing transmission locations and the UAV's trajectory, the system effectively reduces the Age of Information (AoI) and enhances data relevance.