Resource Allocation for Multi-UAV Assisted IoT Networks: A Deep Reinforcement Learning Approach

The wireless communication system for the massively heterogeneous Internet of Things (IoT) network hinders the allocation of resources. For this study, an unmanned aerial vehicle (UAV) is considered as a base station (BS) for IoT system communications to maximize the utilization of future wireless network capacity. Two UAVs are used as the power source and an information transmitter (Tx). To be easily managed, IoT devices are categorized into two clusters of urban and semi-urban features based on their signal distribution and fluctuations. Next, the deep reinforcement learning (DRL) approach is proposed as a resource allocation (RA) scheme. The allocation of resources for bandwidth, throughput, and power consumption issues are considered to be our resources. TensorFlow (Python) programming tool is executed to evaluate and estimate the overall capability of the system. Finally, we analyzed the proposed approach based on different scenarios. Based on the experimental results, our method shows promising outcomes with rapid convergence, suitable for heterogeneous networks, and low complexity based on the evaluation tasks of classification and regression.