Artificial Intelligence-Aid OTFS Receiver: Design and Performance Evaluation

OTFS technology is attracting much attention due to its strong adaptability to high-speed mobility and high spectrum efficiency. Studies show that applying deep learning (DL) to traditional communication systems significantly improves their performance. To this end, this paper proposes an AI-aid OTFS receiver to enhance the performance of OTFS systems in high-speed mobile communication scenarios. The proposed receiver employs two networks - a channel estimation network called TFNet and a decoding network called RecovNet. In the channel estimation module, TFNet performs more accurate channel estimation based on the results of traditional channel estimation methods. In the decoding module, RecovNet decodes the signal based on the equalized signals. To validate the superiority of our proposed receiver, we conduct comparative testing with a traditional receiver under the integer Doppler channel. The experimental results demonstrate that integrating DL technology in OTFS systems effectively enhances their performance.