Learning Meets Logic: Fuzzy Logic Based Deep Matching for Intelligent User Association in 6G IoT Networks

The sixth-generation (6G) wireless communication system enables advanced applications while simultaneously presenting new challenges in fulfilling the demands of these emerging applications, such as high data rates, low latency, high reliability, and massive connectivity requirements. The integration of terrestrial and non-terrestrial networks (TN-NTNs) is considered a promising communication paradigm for 6G wireless communication systems. This combination offers seamless coverage and adaptable wireless access under the “Anywhere, Anytime” concept. However, this integrated framework faces challenges such as limited spectrum availability, unstable and dynamic links, constrained energy resources, and frequently changing network topology. Moreover, the complexity of resource allocation and user association significantly impacts user quality of experience (QoE) and the quality of service (QoS) perceived by users. To address these challenges, we investigate an uplink non-orthogonal multiple access (NOMA)-based TN-NTN system to improve users’ quality of experience by jointly optimizing resource allocation and user association with one of the receivers, including satellite, unmanned aerial vehicles, and a base station. The system employs a hybrid user-receiver association framework by integrating Siamese Neural Networks (SNN) with fuzzy logic reasoning. The SNN model evaluates the features specific to each user-receiver pair, such as transmit power, channel conditions, and subchannel allocation, to determine similarity scores. To refine the similarity score and align the association decision with the user’s application-level requirements, we incorporated a fuzzy logic module that interprets contextual parameters (e.g., application intent and battery level) to ensure that the final association reflects both the user’s intent and environmental compatibility. The numerical results demonstrate that the proposed fuzzy-logic-based SNN approach significantly outperforms the benchmark schemes in terms of latency, reliability, user satisfaction, and energy efficiency, especially in environments with varying QoS requirements and high density.