In recent years, Diffusion Models have emerged as a core technology in the field of generative AI. By progressively adding and removing noise, they exhibit many similarities with signal processing in wireless communications, enabling controllable and high-quality data generation. Based on this, this report explores three key directions: First, leveraging their denoising paradigm, diffusion models are designed as a new module for the physical layer of wireless communications to efficiently eliminate interference and noise, aiming to develop robust wireless transmission methods. Second, utilizing their content generation capabilities, diffusion models are integrated with wireless transmission to extend communication from content decoding and reconstruction to content generation, thus exploring generative communication methods. Finally, taking advantage of their powerful ability to fit complex data distributions, diffusion models are combined with reinforcement learning to form a multi-objective optimization algorithm, which can be applied to wireless communication resource allocation and optimization.