Information Loss Minimization in FANET-Aided Rechargeable IoT Networks Using Q-Learning

The evolution of the Internet of Things (IoT) networks has witnessed tremendous growth in the sensing and monitoring of different environmental activities. As a result, an immense amount of data is processed by low-powered IoT nodes. Moreover, the limited onboard storage on these IoT nodes can result in the loss of valuable information. Recently, Unmanned Aerial Vehicles (UAVs) have been utilized as efficient relaying units for data transmission in IoT networks. A swarm of such UAVs, commonly known as Flying Adhoc Networks (FANETs), is capable of real-time data accumulation and communication in several IoT applications like remote healthcare, crowd surveillance, etc. However, the selection of relay UAVs in FANETs is a challenging issue in FANET-aided data aggregation in rechargeable IoT networks. In this regard, the design of efficient data routing strategies in FANETs for improving the quality of services in these IoT networks is crucial. Adhering to the primary objective of this work, a mathematical problem to jointly minimize the total energy consumed and total information loss in the IoT network is first formulated. Subsequently, a Markov Decision Process (MDP) is developed to illustrate this objective in IoT networks. Moreover, we propose a robust Q-Learning-based Data Forwarding Scheme (QDFS) to generate a reliable data relaying route between source and destination. A multi-parameter reward function is utilized for a systematic selection of relay UAVs. It is observed that QDFS improves the decision-making capabilities of the RL agent, resulting in a suitable choice of reliable relay UAVs, thereby outperforming other routing algorithms. The simulation results motivate the usage of QDFS in FANET-aided self-sustaining IoT networks.