Wireless channel feedback method assisted by environmental knowledge based on AI

The invention discloses an AI-based environment knowledge-assisted wireless channel feedback method, and the method comprises the following steps: (1) quantitatively feeding back downlink channel state information according to a codebook-based feedback architecture of an existing system; (2) collecting a large amount of downlink channel data h and codebook output (shown in the description) corresponding to the downlink channel data h; (3) constructing a neural network which takes the codebook output (shown in the description) as the input and the h as the output; (4) training the neural network built in the step (3) by using the data collected in the step (2); (5) feeding back an index of a code word by a user according to an original codebook, searching a corresponding code word by a base station according to the index, inputting the code word into the neural network trained in the step (4), and outputting a high-precision downlink channel (shown in the description). According to the invention, environmental