Linear Attention Based Channel Estimation Scheme for V2X Communications

In the Vehicle-to-Everything (V2X) environments, the wireless channel exhibits doubly selective fading and high variability, making accurate channel estimation crucial for communication systems. The IEEE 802.11p standard, which has a limited number of pilot symbols, further restricts the accurate capture of channel state variations. This limitation leads to a significant performance degradation of traditional estimators under this standard. Moreover, estimators based on deep learning (DL) still underperform in doubly selective channels or suffer from excessive computational complexity. In this paper, we propose an estimator based on the Receptance Weighted Key Value (RWKV) network with linear attention to capture the correlation between input symbols, combined with data-pilot aided (DPA) estimation to dynamically track channel variations, thereby enhancing estimation performance. Simulation results demonstrate that the proposed RWKV-DPA estimator outperforms other DL-based estimators in terms of performance while maintaining lower complexity.