Efficient traffic flow prediction via bio-inspired reservoir computing with coupled neural dynamics

Abstract

Accurate urban traffic flow prediction is essential for intelligent transportation systems. Traditional time series models and conventional recurrent neural networks (RNNs) often struggle to capture complex nonlinear and long-term temporal dependencies while maintaining computational efficiency. To address this issue, this paper proposes an improved reservoir computing model, termed I-ICM-RC, in which a Simplified Continuous Coupled Neural Network (SCCNN) is employed as the reservoir module. By incorporating bio-inspired integrate-and-fire dynamics and structured local coupling, the proposed model enhances the representation of spatiotemporal patterns in traffic flow. Experiments conducted on real-world traffic datasets evaluate the proposed method under multi-step forecasting scenarios. The results show that the proposed model generally achieves better performance than the standard Echo State Network (ESN), particularly in short-and medium-term prediction tasks, while maintaining low computational cost. These findings indicate that the proposed approach provides an efficient and robust alternative for traffic flow prediction.