Title:A Machine Learning based Hybrid Receiver for 5G NR PRACH

Abstract:Random Access is a critical procedure using which a User Equipment (UE) identifies itself to a Base Station (BS). Random Access starts with the UE transmitting a random preamble on the Physical Random Access Channel (PRACH). In a conventional BS receiver, the UE's specific preamble is identified by correlation with all the possible preambles. The PRACH signal is also used to estimate the timing advance which is induced by propagation delay. Correlation-based receivers suffer from false peaks and missed detection in scenarios dominated by high fading and low signal-to-noise ratio. This paper describes the design of a hybrid receiver that consists of an AI/ML model for preamble detection followed by conventional peak detection for the Timing Advance estimation. The proposed receiver combines the Power Delay Profiles of correlation windows across multiple antennas and uses the combination as input to a Neural Network model. The model predicts the presence or absence of a user in a particular preamble window, after which the timing advance is estimated by peak detection. Results show superior performance of the hybrid receiver compared to conventional receivers both for simulated and real hardware-captured datasets.