A machine learning-based design of PRACH receiver in 5G

The physical random access channel (PRACH) in the uplink of cellular systems is used for the initial access requests from users. In fifth generation (5G) systems three different types of services are available, which are massive machine-type communication, enhanced mobile broadband communication, and ultra-reliable low-latency communication. Considering the tight requirements in terms of latency, a robust design of PRACH receiver is one of the priorities. In this paper we first explore the simple extension of a technique proposed for fourth generation (4G) systems to 5G. Then we propose the application of machine learning techniques to make the PRACH receiver more robust to false peaks, which are responsible of performance degradation in the extension of the 4G technique to 5G. Monte Carlo simulations are used to evaluate and compare the performance of the proposed algorithms.