Deep Learning-Based Self-Organizing Network for B5G Communications with Massive IoT Devices

Self-organizing network (SON) algorithms that were designed for the self-configuration, self-optimization, and self-healing of today's 4G networks exhibit various drawbacks. The two most severe drawbacks are (1) so-called SON use case coordination - the coordination of conflicting network parameter changes - which can lead to sub-optimal network configurations and, more likely, to a worsening of network performance, and (2) a qualitative and quantitative lack of input information to the SON, making reliable network management cumbersome. As a result, SON has been reactive in 4G, which means problem resolution rather than problem prevention, and it has not yet evolved to fully autonomous network optimization as aimed for beyond 5G (B5G) networks. This project focuses on exploratory research on the identification and exploitation of repetitive patterns, auto- and cross-correlations, and intrinsic structures of small- to large-scale spatio-temporal changing wireless conditions, the data throughput experienced by different massive IoT devices, and the (user) data generated, processed, transmitted, and received by these devices (namely Big Data). Deep learning techniques will be used for the identification and a novel machine-learning-based autonomous decision maker (ADM) for the exploitation of the desired information. The goal is to devise a novel deep-learning assisted ADM, using algorithms for the allocation of radio and network resources in a B5G network servicing massive IoT and serveing as an input to legacy SON. Hence, a novel multi-layer SON architecture will be introduced to fully exploit the flexibility of B5G networks.
 

Funding for this research provided by the National Science Foundation.