ML-Based Receiver Design for Chirp-Based Asynchronous Pulse Code Multiple Access for Massive IoT

Asynchronous Pulse Code Multiple Access (APCMA) is a pulse-based wireless communication scheme designed for Massive IoT, offering high resilience against largescale packet transmissions from numerous devices. To enable longrange communication similar to other LPWA technologies, Chirp Spread Spectrum (CSS) has been incorporated into APCMA. However, conventional threshold-based detection methods suffer from critical limitations in low-SNR environments, where optimal threshold tuning and accurate signal detection become challenging. In this paper, we propose a machine learning-based pulse detection method to enhance the reception sensitivity of APCMA. The proposed approach is implemented using USRP and GNU Radio, and its performance is evaluated through 920 MHz band wireless communication experiments under low SNR and high-density conditions. The results demonstrate that, compared to conventional APCMA, the proposed method eliminates the need for manual threshold configuration and significantly improves detection performance.