A guidance assistance strategy for intersections based on vehicle speed prediction sequences

To improve the efficiency of urban traffic systems and environments, a speed guidance model based on a predictive vehicle speed sequence is proposed in this study. The model introduces a multi-step model of vehicle speed prediction based on nonlinear autoregressive models with a multi-source exogenous inputs (NARXs) neural network, with fusion of multi-source exogenous inputs into the speed guidance model to assist the vehicles in efficiently passing the intersections. The implementation of traffic flow simulation of urban road networks was established using SUMO-Python software, and experiments were carried out from the two dimensions of different speed guidance models and traffic networks with different numbers of lanes. The results demonstrate that the guidance accuracy of the proposed guidance strategy is higher than that of the previous vehicle speed guidance strategy, which can be improved by 16.5 %, significantly reducing the idle time of vehicles at intersections and improving urban traffic safety and operational efficiency. Moreover, the proposed speed guidance strategy is applicable to different urban road networks, which can reduce fuel consumption and pollutant emissions to a greater extent, efficiently improving the economy, sustainability, and operational environment of urban traffic.