Adaptive Task Scheduling via Graph Convolutional Deep Q‐Networks for Dynamic Dependency Modeling

Task scheduling in dynamic environments poses substantial challenges due to the intricate interdependencies among tasks and the continuously evolving system states. In this paper, we introduce an ingenious Adaptive Graph‐Convolutional Deep Q‐Network (AGC‐DQN) framework that seamlessly integrates Graph Neural Networks (GNNs) with the Deep Q‐Networks (DQN) paradigm to effectively address these challenges. By representing task attributes and their dependencies as graph structures, our approach captures the complex relationships between tasks and adaptively computes feature interactions via graph convolution operations. This architecture enables the reinforcement learning agent to make more informed, context‐aware scheduling decisions, enhancing its decision‐making capabilities. We evaluate the AGC‐DQN framework on the benchmark dataset and the dynamic scheduling scenario, demonstrating its superior performance in terms of task completion rate, system efficiency, and adaptability compared to traditional DQN‐based methods. Our results underscore the potential of using graph‐based representations to improve task scheduling in complex and dynamic environments. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.