Adaptive preprocessing degree-guided joint optimization of resources and task result confidence in Industrial IoT edge computing

With the rapid growth in the deployment scale of Industrial Internet of Things (IIoT) devices, the continuous generation of massive computational tasks poses severe challenges to the limited resources at the edge. Data preprocessing can effectively reduce transmission load and save bandwidth resources, but excessive preprocessing degrees can lead to the loss of critical information, thereby reducing task result confidence. To address this problem, this paper proposes a deep reinforcement learning-based joint optimization method for resource allocation and task result confidence. First, a resource cascade conduction relationship model in multi-layer heterogeneous networks is constructed, laying a theoretical foundation for cross-layer collaborative optimization. Second, with preprocessing degree as the core decision variable, the coupling constraint relationships among preprocessing degree, resources, task result confidence, and energy consumption are established. Furthermore, the Cascade Feature Encoding Dueling Deep Q-Network (CFE-Dueling DQN) algorithm is designed based on the resource conduction mechanism to solve multi-constraint optimization problems aimed at maximizing the comprehensive utility of the system. Simulation experimental results show that the proposed method demonstrates significant performance advantages over baseline strategies under different numbers of devices and different collaboration scale constraints in industrial scenarios.