@

초록

This paper addresses the broadcast scheduling problem in multi-hop wireless sensor networks, a critical challenge that affects the efficiency and reliability of various network operations. We propose to transform the broadcast scheduling problem into a node classification problem, in which the proposed model predicts the receiving time slot of each sensor node. Utilizing graph convolutional networks to capture dependencies between nodes in the broadcasting structure, our model takes the hop distance to the sink, number of children, and number of descendants as the input features. To make training more efficient, we improved the model training by using a teacher-student framework, where we combine the losses from both the teacher and student models for learning. Our results show the effectiveness of this approach, with the predicted transmission schedule achieved up to 99% accuracy.