Energy Efficient Image Transmission Over Semantic Communication Enabled UAV Swarm Networks

In this paper, a semantic communication framework is proposed for unmanned aerial vehicle (UAV) swarm formation flight. In the studied model, a leader extracts semantic information from the aerial image and transmits it to each follower using limited transmission power. The borders of obstacles are referred to semantic information of the aerial image. Based on the received semantic information, each follower reconstructs semantic maps for trajectory planning so as to maintain formation and avoid obstacles. We use the metric of Intersection over Union (IoU) to evaluate the effectiveness of the proposed semantic communication network. The problem is formulated as an optimization problem whose goal is to maximize the sum of IoUs of all followers. To solve this problem, a semantic information selection and power allocation (SISPA) algorithm based on reinforcement learning (RL) framework is proposed. The proposed SISPA algorithm enables the leader to evaluate the importance of semantic information, and, thus efficiently allocating transmission power and selecting important semantic information to improve the sum of IoUs. Simulation results demonstrate that, compared with the traditional wireless communication networks without semantic communication techniques, the proposed approach yields a 17.4% improvement in terms of the sum of IoUs.