In an ultradense Industrial Internet of Things (UDI-IoT) network with device-to-device (D2D) communication technology applied, interaction among a large number of Industrial Internet of Things devices (IIoTDs) leads to heavily overlapping interference, and Age-of-Information (AoI) sensitive service in the network is difficult to guarantee. In this article, a learning-based robust resource allocation considering overlapping interference and AoI-sensitive services requirements (LRRA-OIAoISR) is proposed to solve the problem of AoI resource management for the UDI-IoT networks with overlapping interference. We first construct the interference hypergraph model and analyze the interference relationship between D2D devices, which can effectively improve the utilization of spectrum resources. Then, an AoI model was constructed to address the resource management issues of AoI-sensitive services, and a robust optimization model was established under imperfect channel state information (CSI). This model considers power control and resource conflict constraints to achieve maximum network throughput. Finally, to solve this robust optimization problem, we propose an LRRA-OIAoISR algorithm based on learning theory. The algorithm obtains an optimal robust optimization solution by reducing the impact of imperfect CSI. The performance of the proposed resource allocation method was verified through simulation. Compared with other benchmark algorithms, the proposed algorithm improves energy efficiency by an average of 52.5%, interference efficiency by an average of 88.5%, and throughput by an average of 125%.