Uncertainty-resilient control of large-scale industrial IoT networks via learning-guided fuzzy intelligence

Industrial IoT networks operate under persistent uncertainty arising from node mobility, energy depletion, and dynamic link conditions, which significantly affect communication reliability and network lifetime in large-scale deployments. Existing learning-based and fuzzy routing approaches primarily optimize instantaneous performance metrics but lack a systematic mechanism for maintaining consistent performance under uncertainty. This work proposes a learning-guided fuzzy control framework for uncertainty-aware IIoT communication, integrating an Interval Type-2 Fuzzy Inference System with reinforcement learning-based clustering. The framework models cluster formation and multi-hop routing as a closed-loop adaptive process, where fuzzy rule strengths are continuously updated through Q-learning feedback. Rather than enforcing strict control-theoretic stability, the proposed approach aims to maintain an empirically stable network behavior around a lifetime-reliability operating region, defined by bounded packet loss, sustained reliability levels, and controlled energy consumption under dynamic conditions. This operating region is achieved by minimizing uncertainty amplification across routing paths through adaptive decision-making. Extensive NS-3 simulations conducted in Factory Hall and Local Automation Unit scenarios demonstrate that the proposed framework achieves consistent performance, including reliability above 96%, reduced packet loss, and up to 9% improvement in network lifetime compared with recent fuzzy-learning baselines.