Prediction of Transit Time on Urban Roads Based on Particle Filtering

Received: 10 December 2019 Accepted: 21 February 2020 The transit time on roads is essential to the intelligent traffic system in urban areas. If the transit time is predicted accurately, the traffic management system can work more effectively, and the public will enjoy rational travel strategies. However, the traditional prediction algorithms for transit time cannot adapt well to the complex road network and the sparsely deployed sensors in cities. By contrast, the particle filtering (PF) algorithm has strong adaptability to such a stochastic nonlinear problem. Therefore, this paper measures the spatiotemporal similarity between historical data on different roads at different moments with speed matrices, aiming to prevent the data degradation in transit time prediction. Besides, the PF algorithm was innovatively applied to build a transit time prediction model on urban roads. The traffic trend in historical data was modelled with weighted particles. Finally, the effectiveness of our algorithm was demonstrated through an empirical analysis. The results show that our algorithm outperforms the other transit time prediction algorithms in prediction accuracy and computing performance.