A method for controlling crowd congestion based on spatial-temporal graph neural network and improved artificial bee colony algorithm

This paper proposes a novel crowd congestion control method that combines spatial-temporal graph neural network with improved artificial bee colony algorithm to better integrate congestion prediction with path regulation in evacuation scenarios. First, crowd data is collected via smartphones using the PDR algorithm, and a real-time knowledge graph (RTKG) is constructed to represent the evacuation scenario. Spatial and temporal features are extracted using a two-layer graph convolutional network (TL-GCN) and a gated recurrent unit with attention (A-GRU) to predict congestion levels. These predictions are integrated into the RTKG. The INFABC algorithm then guides path planning based on these predictions. In this paper, we propose a crowd congestion control system and apply it to crowd evacuation. Experimental results demonstrate that our proposed method enhances the efficiency of crowd evacuation and offers crucial guidance for managing large-scale crowds.