Trade-off of cognition and semantic communication for UAV-IRS enhanced intelligent networks

This paper investigates the trade-off between cognition and semantic communication in a wideband cognitive radio (CR) semantic communication system, where an intelligent reflecting surface (IRS) mounted on an unmanned aerial vehicle (UAV) enhances spectrum sharing through spectrum sensing. In this system, the cognitive base station transmits semantic symbols to users over available spectrum resources, with system performance closely tied to the quality of semantic information delivery. The UAV-IRS integration dynamically improves channel conditions during both sensing and transmission phases, facilitating the stable and efficient delivery of semantic content. However, combining spectrum sensing and dynamic decision-making with the high-throughput, low-latency requirements of modern semantic communication systems remains a challenging task. To improve the efficiency of semantic transmission and optimize the resource allocation between cognition and communication, this paper proposes a joint optimization framework tailored for semantic-aware communication. The system jointly optimizes sensing time, UAV trajectories, beamforming, IRS reflection coefficients, and subcarrier allocation to maximize the overall semantic transmission rate, thereby supporting robust and efficient semantic communication under spectrum-constrained environments. Given the non-convex nature of the optimization problem, we employ a combined D3QN-DDPG algorithm. This hybrid approach leverages the strengths of both Q-learning and DDPG, enabling efficient handling of the complex, high-dimensional state and action spaces. Simulation results show that the proposed joint optimization method effectively balances cognition and communication performance in the UAV-IRS enhanced wideband CR system. Moreover, compared to the baseline scheme, the proposed approach achieves significant improvements in both system capacity and communication rate, demonstrating the effectiveness of the optimization strategy in practical deployment scenarios.