关键词: 短时交通流, 预测模型, Volterra级数, 相空间重构, 混沌

Abstract: Based on phase space reconstruction of chaos dynamic system and Volterra series for nonlinear system, Volterra adaptive prediction model for traffic flow is constructed. On the premise that small data set method is used to determine that chaos exists in traffic flow time series, this paper respectively employs average mutual information method and false nearest neighbor technique to choose delay time and embedding dimension so as to perform phase space reconstruction for traffic flow data. Nonlinear prediction model, whose coefficients are real-time updated by LMS adaptive algorithm is constructed by applying Volterra series extensions. It applies this Volterra prediction model to performing simulations for the real measured expressway traffic flow data and chaotic time series generated by Chens and Duffing. Experimental results show that the proposed Volterra adaptive prediction model is capable of effectively predicting traffic flow time sequence and low-dimensional chaotic time sequence.

Key words: short-term traffic flow, prediction model, Volterra series, phase space reconstructio