Deep Learning-Based Joint Pilot Location and Number Optimization for Channel Estimation in OFDM Systems

In this paper, we propose a novel deep neural network (DNN)-based pilot design scheme that jointly optimizes the number and locations of pilot symbols in closed-loop orthogonal frequency-division multiplexing (OFDM) system. The proposed scheme consists of an encoder and a decoder network. The encoder adaptively determines both the optimal number and the locations of pilot symbols for channel estimation, based on the current channel conditions. The decoder then performs channel estimation using the pilot symbols determined by the encoder. Simulation results show that the proposed scheme outperforms the conventional channel estimation method, while requiring fewer pilot signals on average. Also, when compared to a state-of-the-art deep learning (DL) approach that focuses only on pilot locations, the proposed scheme shows robust channel estimation performance across a diverse range of multipath propagation scenarios.