In this paper, we present a codebook design with deep reinforcement learning (DRL) for orthogonal frequency division multiplexing with all index modulation (OFDM-AIM). The prior AIM codebook design usually has poor error performance at low SNR, and the complexity of finding the optimal codebook through exhaustive search is NP-hard. Therefore, we use DRL to reduce complexity and achieve better error performance. We map the problem of designing codebooks to DRL and design the reward function and other elements. We use the deep recurrent Q-network to make it applicable to codebooks with longer codeword lengths. We analyze the relationship between the Hamming distance and Euclidean distance of the optimal codebook and derive the upper bound of the BER performance. The BER performance is close to the theoretical upper bound and outperform the classic AIM. Through numerical analysis, we find that increasing the number of subcarriers can improve the error performance under the same spectrum efficiency. However, prior exhaustive search codebook design methods are impractical due to the high complexity, which makes the learning approach attractive.
Codebook Design of All Index Modulation with Deep Reinforcement Learning
2022-09-01
1874936 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Codebook Based Precoding for Spatial Modulation
IEEE | 2019
|Springer Verlag | 2012
|BASE | 2019
|