Semantics-Aware Multi-UAV Cooperation for Age-Optimal Data Collection: An Adaptive Communication based MARL Approach

Due to the superior flexibility and extensive coverage, multiple unmanned aerial vehicles (UAVs) cooperation is a promising approach for data collection in improving the information freshness. In this paper, we consider a multi-UAVassisted Internet of Things (IoT) network, where UAVs are deployed to collect data from sensor nodes (SNs) and transmit data back to the BS via wireless links so as to improve the information freshness, measured by the age of information (AoI). It is of great challenge to achieve effective cooperation under distributed decision-making because of the time-varying and stochasticity of the environment and the limited communication range of UAVs. To address this issue, we formulate the problem of joint trajectory plan, SN scheduling, and transmission scheduling as a decentralized partially observable markov decision process (Dec-POMDP), and develop an adaptive communication based multi-agent deep reinforcement learning (AC-MARL) algorithm to solve it. By applying our proposed AC-MARL algorithm, a more effective cooperation can be achieved by exploiting the benefits of semantic-aware communications among UAVs.