Key points are not available for this paper at this time.
As with any other wireless technology, 5G is not immune to jamming. To achieve consistent performance, network resource scheduling must be optimized in a way that reacts to jamming in the NR channel environment. This paper presents a cognitive system for real-time Channel Awareness and Radio Access Network (RAN) Scheduling (CARS) optimization based on multi-dimensional temporal machine learning models. Our system automatically detects and classifies jamming in the channel environment and optimizes scheduling based on classification results and collected link parameters. Based on over-the-air (OTA) experiments, detection and classification time is less than 0.8 seconds, which enables real-time optimization. The system is evaluated and verified for OTA experimentation through integration to our end-to-end NR system. An Automated Jamming Module (AJM) is designed and implemented. Connecting the AJM to our NR system enables a comprehensive evaluation environment for our Jamming Detection and Classification Model (JDCM) and Modulation and Coding Scheme optimization model. The improvement in connection resiliency against Control Resource Set jamming is proof of the CARS concept for real-time channel awareness and scheduling optimization. Depending on channel conditions, CARS achieves a 30% or higher improvement in NR system throughput.
Wang et al. (Mon,) studied this question.