ML-Based Preamble Collision Detection in the Random Access Procedure of Cellular IoT Networks

TL;DR

This paper introduces a machine learning-based method for early collision detection in the random access procedure of cellular IoT networks, demonstrating high accuracy and low latency suitable for real-time deployment.

Contribution

It presents a novel ML-based collision detection mechanism, evaluates multiple classifiers under realistic conditions, and optimizes the model for deployment on base station hardware.

Findings

Neural network achieved over 98% accuracy in in-distribution scenarios.

Quantization reduced inference time from 2500 ms to 0.3 ms with minimal accuracy loss.

The method is suitable for real-time, scalable IoT applications.

Abstract

Preamble collision in the random access channel (RACH) is a major bottleneck in massive machine-type communication (mMTC) scenarios, typical of cellular IoT (CIoT) deployments. This work proposes a machine learning-based mechanism for early collision detection during the random access (RA) procedure. A labeled dataset was generated using the RA procedure messages exchanged between the users and the base station under realistic channel conditions, simulated in MATLAB. We evaluate nine classic classifiers -- including tree ensembles, support vector machines, and neural networks -- across four communication scenarios, varying both channel characteristics (e.g., Doppler spread, multipath) and the cell coverage radius, to emulate realistic propagation, mobility, and spatial conditions. The neural network outperformed all other models, achieving over 98\% balanced accuracy in the in-distribution evaluation (train and test drawn from the same dataset) and sustaining 95\% under out-of-distribution evaluation (train/test from different datasets). To enable deployment on typical base station hardware, we apply post-training quantization. Full integer quantization reduced inference time from 2500 ms to as low as 0.3 ms with negligible accuracy loss. The proposed solution combines high detection accuracy with low-latency inference, making it suitable for scalable, real-time CIoT applications found in real networks.