Abstract
Emerging low-altitude wireless networks (LAWNs) provide a promising aerial platform for integrated sensing and communication (ISAC) due to the high mobility and flexible deployment. Nevertheless, dynamic environment and limited infrastructure support pose significant challenges for achieving high-accuracy ISAC services. To address these issues, we propose a Transformer-based integrated sensing and semantic communication (T-ISSC) framework that transmits task-relevant semantic tokens in distributed sensing networks. In particular, inspired by recent advances in generative large models, the proposed T-ISSC framework integrates the Transformer-empowered joint source-channel coding architecture with a task head for target state estimation. Moreover, transfer learning is leveraged to enhance the framework's adaptability to diverse communication environments in LAWNs. Simulation results demonstrate that the proposed T-ISSC framework achieves a 52.35%-73.66% reduction in root mean square error compared with benchmark schemes.