RACH Traffic Prediction in Massive Machine Type Communications

TL;DR

The paper tackles bursty traffic prediction in massive mMTC networks by proposing a lightweight, online ML framework that combines an LSTM and DenseNet (with residuals) for traffic prediction and a burst-detection network for congestion forecasting. A key novelty is the Fast LiveStream Predictor (FLSP), which updates LSTM states using only fresh data, avoiding buffering and reducing computation compared to traditional rolling methods, while preserving long-range dependencies. The framework is evaluated in a realistic, multi-group mMTC scenario, demonstrating substantial improvements in long-term prediction accuracy (up to ~52% reported) and superior burst detection performance (high F1-scores) with lower processing load. These results support proactive, time-critical congestion management in large-scale IoT networks and offer a flexible methodology adaptable to various recurrent architectures and live-time-series applications.

Abstract

Traffic pattern prediction has emerged as a promising approach for efficiently managing and mitigating the impacts of event-driven bursty traffic in massive machine-type communication (mMTC) networks. However, achieving accurate predictions of bursty traffic remains a non-trivial task due to the inherent randomness of events, and these challenges intensify within live network environments. Consequently, there is a compelling imperative to design a lightweight and agile framework capable of assimilating continuously collected data from the network and accurately forecasting bursty traffic in mMTC networks. This paper addresses these challenges by presenting a machine learning-based framework tailored for forecasting bursty traffic in multi-channel slotted ALOHA networks. The proposed machine learning network comprises long-term short-term memory (LSTM) and a DenseNet with feed-forward neural network (FFNN) layers, where the residual connections enhance the training ability of the machine learning network in capturing complicated patterns. Furthermore, we develop a new low-complexity online prediction algorithm that updates the states of the LSTM network by leveraging frequently collected data from the mMTC network. Simulation results and complexity analysis demonstrate the superiority of our proposed algorithm in terms of both accuracy and complexity, making it well-suited for time-critical live scenarios. We evaluate the performance of the proposed framework in a network with a single base station and thousands of devices organized into groups with distinct traffic-generating characteristics. Comprehensive evaluations and simulations indicate that our proposed machine learning approach achieves a remarkable $52\%$ higher accuracy in long-term predictions compared to traditional methods, without imposing additional processing load on the system.

Figures (12)

• Figure 1: An example of application-based categorization of devices in an mMTC network. The total network traffic is the sum of the traffic generated by the various groups, each with their own set of characteristics.
• Figure 2: Four-way handshake procedure of access request in LTE and NR-5G.
• Figure 3: Proposed ML architectures for traffic prediction and congested bursty event detection networks. Second network receives the output of the first network as chunks of data.
• Figure 4: Recursive algorithm in the prediction phase of LSTM networks.
• Figure 5: Comparison of rolling and FLSP algorithms in generating output sequences. (a) Rolling algorithm: The input sequence at each step (solid black frame) consists of truncated historical data and fresh data (blue shade). The traffic prediction network generates predictions for $l_p$ time slots at each step (solid red frame), but only the last $l_f$ time slots (green shade) are retained and appended to the output sequence. (b) FLSP algorithm: In this method, the input sequence includes only the fresh data, resulting in faster state updates compared to the rolling method. Yellow dashed lines represent a chunk of data, which is the concatenation of fresh data and predicted samples at each step, used as input for the burst detection network.