Unmanned aerial vehicles (UAVs) have been considered a promising technology in the operation of Internet of Things (IoT) networks. IoT networks play a significant role in different applications that allow us to connect, collect, and process data. Yet,...
Unmanned aerial vehicles (UAVs) have been considered a promising technology in the operation of Internet of Things (IoT) networks. IoT networks play a significant role in different applications that allow us to connect, collect, and process data. Yet, effective data collection and processing in IoT networks remain one of the key concerns since IoTDs are resource-constrained, the working environment is dynamic, and data delivery must be timely. These challenges have triggered to propose a MEC-enabled UAV-assisted IoT network system where UAVs are equipped with MEC functionalities to make the data collection and process rewarding. In this paper, we study how the trajectory of UAVs should be optimized to minimize the age of information (AoI) and mission time. We apply the K-means dynamic clustering algorithm to cluster IoTDs based on their geographical location proximity, which helps to simplify the trajectory optimization problem. Then, we develop a novel Deep Deterministic Policy Gradient-based Dynamic Hovering Point Selection Algorithm (DDPG-DHPSA) to identify the hovering point within each cluster. The hovering point is selected by the DDPG-DHPSA based on the IoTD with the highest data transmission time and AoI to acquire the necessary data quickly. We incorporate sparse code multiple access (SCMA) as a communication model to improve the capability of data transmission between the UAV and IoTDs. The optimization problem is formulated as a Markov decision process (MDP) to control the continuous decision-making nature of the UAV’s actions. DDPG-DHPSA adopts deep neural networks as function approximators to handle the high-dimensional, continuous state and action spaces associated with UAV trajectory optimization. Extensive simulations are conducted to evaluate the performance of the proposed solutions. The results clearly show high performance with the proposed K-means dynamic clustering, DDPG-DHPSA combined with SCMA to minimize AoI and mission time in contrast to the baseline scenarios. By employing K-means dynamic clustering, the system’s performance improved by approximately 87.5%, while the implementation of the DDPG-DHPSA algorithm yields an improvement of about 73%.