Semantic-aware communication and computation co-optimization in digital twin-assisted Internet of Drones networks using federated deep reinforcement learning

This paper introduces a semantic-aware framework for communication and computation in Internet of Drones (IoD)-assisted digital twin (DT) networks within the context of smart agriculture. In our system, the IoD gathers environmental data and conducts onboard semantic analysis to identify task-relevant features before transmission. We propose a joint optimization scheme for offloading and semantic processing that utilizes federated deep reinforcement learning (FDRL) to address communication and computation constraints via context-adaptive semantic communication. The simulation results show that the proposed approach substantially enhances energy efficiency and communication delay, depending upon the configuration, compared to traditional offloading strategies.