Physics-Informed Uncertainty-Aware IoT Networking for Dense Urban and Urban Micro Environments

Large-scale IoT sensor networks operating in dense urban and urban microcell environments encounter challenges due to high node density, strong co-channel interference, complex building-induced propagation effects, and uneven energy consumption across the network. These conditions often limit the effectiveness of static deployment approaches and data-driven optimization methods in interference-heavy smart city scenarios. To address these issues, this work proposes a physics-informed and uncertainty-aware learning framework designed for Urban Micro and Dense Urban IoT deployments. The framework integrates a Physics-Informed Neural Network with a Firefly-based global optimization strategy to enhance deployment efficiency, ensure physical consistency, and improve resilience under uncertain operating conditions. By embedding wireless propagation constraints directly into the learning process, the approach produces physically meaningful predictions even in highly obstructed environments. Additionally, stochastic effects and prediction uncertainty are explicitly modeled, allowing the system to identify unreliable regions and adapt accordingly. Simulation results show improved energy balance, more stable connectivity, and higher packet delivery reliability, demonstrating the framework's suitability for next-generation smart city IoT networks in complex urban settings.