本研究では，無線通信路に適した適応量子化及び量子化情報に基づく信号処理設計と，深層学習に基づくパラメータ最適化・入力推定を用いた超大容量・超低遅延・超低消費電力無線通信の実現を目指す．
まず無線通信システムの様々な処理で重要となる通信路状態情報（CSI）のフィードバック方法について，自動符号化器）と転移学習，量子化等に基づくフィードバック情報量が少なく高い精度が得られる方法を複数提案し，IEEEの国際会議等で発表した．また，データを共有せずに優れた学習モデルを構築できる連合学習について，従来問題となっていたデバイスでの演算量とデバイスーサーバー間の通信量を減らす新たな連合学習を提案した．
We propose a feedback method for CSI, based on autoencoder, trasnfer learning, quantization, and so on. We have proposed several methods to obtain high accuracy with small amount of feedback information, and have been presented at IEEE international conferences, etc. In addition, we proposed a new method of coalition learning that reduces the amount of computation in the device and the amount of communication between the device and the server, which had been a problem in the past. The results were presented at the IEEE International Conference on Wireless Communication Systems, and others. In addition, we proposed a new distance estimation method based on deep learning using Massive MIMO and millimeter wave, which will be key technologies in 5G and Beyond 5G.
We also proposed a new distance estimation method based on deep learning for Massive MIMO and millimeter wave, which are key technologies for 5G and Beyond 5G.

研究種目 : 基盤研究 (B) (一般)
研究期間 : 2019～2022
課題番号 : 19H02142
研究分野 : 無線通信工学