Semantic-Aware UAV Swarm Coordination in the Metaverse: A Reputation-Based Incentive Mechanism

Unmanned aerial vehicle (UAV) swarms have found extensive applications owing to their flexibility, mobility, cost-effectiveness, and capacity for collaborative and autonomous service delivery. Empowered by intelligent algorithms, UAV swarm can exhibit cohesive behaviors and autonomously coordinate to achieve collective objectives. Nonetheless, in real-world scenarios with uncertainty and stochasticity, its performance suffers from the unstable information exchange among UAVs and inefficient data sampling. In this paper, we introduce a metaverse-based UAV swarm system, where monitoring, observation, analysis, and simulation can be realized collaboratively and virtually. Within the metaverse, virtual service providers (VSPs) utilize digital twin (DT) to generate and render virtual sub-worlds, while providing diverse virtual services. In particular, the VSP trains the learning model using high-fidelity data from the physical world, formulates optimal decisions for diverse tasks, and returns these decisions to the UAV swarm for the execution of the corresponding tasks. Since synchronization between two worlds needs frequent data exchange, we employ the semantic communication technique in our system which could reduce communication latency by transmitting only the semantic information. In such design, UAVs as workers are employed to collect data and provide extracted semantic information to the VSPs. Moreover, we propose a hierarchical framework to investigate the reliability and sustainability of the metaverse-based UAV swarm system. In the lower layer, we design a worker selection scheme to determine reliable UAVs for data synchronization. In the upper layer, we consider deep learning (DL)-based auction as the incentive mechanism for resource allocation in semantic information trading between UAV swarm and VSPs.