道路交通流状态的多参数融合预测方法

Multi-parameter fusion applied to road traffic condition forecasting

针对道路交通流普遍存在的混沌特性以及单交通参数不足以全面反映交通流状态的实际情况,考虑交通动力学系统中多个交通参数之间的关联关系,提出一种新的多参数混沌时间序列预测算法.该算法在相空间重构理论的基础上,借助Bayes估计将多个参数在同一高维相空间中进行相点最优融合,从而增加重构相空间的系统信息量,使得相空间的相点轨迹更加逼近原交通系统的动力学行为.同时借鉴单参数混沌时间序列预测方法,从不同角度对动力学系统的运动状态进行描述,以实现多参数时间序列的混沌预测.实验结果表明,通过融合多交通参数时间序列,获得了更加完整的交通流状态变化特征.与单交通参数时间序列的预测结果相比,其预测误差显著降低,均衡系数相应增大,提高了交通流状态预测的准确率.

In view of the chaotic characteristic in road traffic flow and the actual traffic condition that cannot be comprehensively reflected by single traffic parameter, a fusion algorithm of multi-parameters for traffic condition forecasting, with the consideration of the relationship between multiple parameters, is proposed. This algorithm is based on the reconstruction of phase space. According to Bayesian estimation theory, the multiple traffic parameters are optimally fused into phase points in the same phase space. Accordingly, the phase space information increases and the phase points are closer to dynamical behavior of the traffic system. On the basis, by using the multi-parameter chaos prediction method, the tendency of dynamic systems from different aspects is described, with reference to the method of predicting single parameter chaotic time series. The experimental results confirm that more features of real traffic condition are reflected by fusing multiple traffic parameters. The multi-parameters forecasting algorithm reduces the prediction error and improves the equalizer coefficients compared with the results generated from single parameter prediction. That is to say, the prediction method used in this paper is effective and accurate for predicting traffic condition based on multiple traffic parameters.