Multi-agent Reinforcement Learning-based Resource Allocation Scheme for UAV-assisted Internet of Remote Things Systems

Multi-layered communication networks including satellites and unmanned aerial vehicles (UAVs) with remote sensing capability are expected to be an essential part of next-generation wireless systems. It has been reported that deep reinforcement learning algorithm brings performance improvement in various practical environments. However, it is anticipated the computational complexity will a critical issue as number devices network significantly increased. To resolve this problem, paper we propose multi-agent (MARL)-based resource allocation scheme for UAV-assisted Internet things (IoRT) The UAV IoRT sensors set MARL agents, which independently trained minimize energy consumption cost by controlling transmit power bandwidth. shown agent can reduce 70.9195 kJ, while sensor agents yield 20.5756 kJ reduction, 65.4 % 71.97 reductions compared initial state each agent. Moreover, effects from hyperparameters neural episodic control (NEC) baseline investigated terms consumption.