Channel Knowledge Map Empowered Low-Overhead Channel Estimation Scheme in MIMO-OFDM Systems

In MIMO-OFDM system, channel estimation is a crucial step for achieving stable data transmission. The base station (BS) can achieve accurate channel estimation by sending pilot signals. However, with the continuous increase in antenna dimensions and frequency band width, the pilot overhead of traditional channel estimation methods has significantly increased, which poses a great challenge to efficient data transmission. Fortunately, the channel knowledge map (CKM) can infer the inherent channel information using only location information, providing a new approach to solve this problem. Based on this idea, this paper realizes channel frequency-spatial extrapolation using CKM under the condition of low pilot overhead, significantly improving the accuracy of channel estimation and laying the foundation for efficient data transmission. Specifically, we first introduce a scenario representation network to construct CKM, enabling CKM to store the frequency-spatial correlation between adjacent channels. Then, we use pilot signals to estimate the channels of a limited number of sub-carriers and antennas, providing the physical location information required by CKM. Finally, we input this location information into CKM to extrapolate the channels in the entire frequency and spatial domains. Numerical results are provided to validate the superiority of the proposed channel estimation method as compared to the other methods in terms of SNR and spectral efficiency, and this superiority becomes more pronounced under low pilot overhead.