Adaptive 6G networks are crucial for maintaining reliable communication in dynamically changing environments. To this end, accurate propagation models are essential for network adaptation, but developing real-time, high-fidelity models is a significant challenge. Machine Learning (ML)-based channel models have gained attention in recent years, yet their limited generalizability often necessitates extensive retraining to ensure applicability in unseen environments. In this paper, we propose a process chain for rapid network adaptation in dynamic environments. Our fully automated process encompasses network Key Performance Indicator collection, LiDAR-based environmental modeling, Transfer Learning (TL)-assisted refinement of the ML-based IndoorDRaGon propagation model, and network planning. We demonstrate that Artificial Neural Networks (ANNs) achieve comparable accuracy to Random Forest (RF)-based models while offering significantly faster adaptation to new environments. Unlike RFs, ANNs do not need to be retrained from scratch using all the source data, which reduces computational effort and improves data privacy as only the new data is required. This enables efficient and dynamic network reconfiguration in previously unseen scenarios with just a few new measurements.

Wireless 5G connectivity in private operation is rapidly evolving into a necessity component for business-critical services within professional industries, such as those related to manufacturing and logistics. The resulting large number of private 5G networks requires reliable methods for network planning and operation, for which current research is already presenting approaches for efficient AI-based automated network planning. However, there is often a significant gap between current research findings and their practical application in industry, resulting in insufficient delays for the utilization of research outcomes by industrial stakeholders. Hence, this paper introduces a solution that provides industry partners with free and open access to the latest research findings on automated network planning. As primary contribution, we present a web-based planning tool facilitating intuitive network planning within a few simple steps. As a result, radio environmental maps are generated using deep learning propagation modeling as well as AI-driven network planning outcomes are evaluated, balancing Quality of Service (QoS) levels against the requisite number of base stations. Video Abstract-Demonstration video is available online at https://tiny.cc/Private5GPlanner

Private networks represent a key innovation in current 5G and future 6G networks, offering significant benefits, particularly for vertical industries with mission-critical industrial applications. Compared to public networks, the deployment of numerous potential private networks demands automated network planning while simultaneously meeting higher performance requirements for targeted applications. Emerging approaches have successfully utilized AI-based methodologies as a basis for automated network planning in greenfield deployments within licensed but purely private frequency bands. However, these approaches fail to include and extend brownfield implementations in public mobile networks, which would be crucial for private networks running a shared operator model. Thus, this paper presents an AI-based automated network planning methodology augmented by our recently introduced and thoroughly validated data-driven channel modeling approach, DRaGon. Further, this combined AI-based planning methodology is extended to provide automated network planning solutions for shared private networks within public macro networks. The overall planning accuracy was successfully validated with only minor deviations using public network deployments as ground truth. As a key result, we demonstrate that the performance of the presented AI-based planning method can reliably and accurately plan demand-driven network expansions for professional applications with the highest quality requirements.

Radio channel modeling is one of the most fundamental aspects in the process of designing, optimizing, and simulating wireless communication networks. In this field, long-established approaches such as analytical channel models and ray tracing techniques represent the de-facto standard methodologies. However, as demonstrated by recent results, there remains an untapped potential to innovate this research field by enriching model-based approaches with machine learning techniques. In this paper, we present Deep RAdio channel modeling from GeOinformatioN (DRaGon) as a novel machine learning-enabled method for automatic generation of Radio Environmental Maps (REMs) from geographical data. For achieving accurate path loss prediction results, DRaGon combines determining features extracted from a three-dimensional model of the radio propagation environment with raw images of the receiver area within a deep learning model. In a comprehensive performance evaluation and validation campaign, we compare the accuracy of the proposed approach with real world measurements, ray tracing analyses, and well-known channel models. It is found that the combination of expert knowledge from the communications domain and the data analysis capabilities of deep learning allows to achieve a significantly higher prediction accuracy than the reference methods.