Abstract: Due to complicacy, randomness and nonlinearity of traffic system, traffic flow parameters are usually considered as the random time series. A method based on G-P algorithm that is helpful to convert the traffic flow time series to stationarity series was proposed. First, the autocorrelation function was adapted to evaluate the stationary of traffic flow under free traffic, congested traffic and jam traffic. Second, G-P algorithm was used to calculate the embedding dimension, reconstruct the phase space, and convert the traffic flow non-stationary parameters to stationary time series under 3 traffic states. At last, some examples illustrated the model and showed its practical applicability based on measured traffic flow data. The research can provide the input set for the traffic flow parameters analysis, fitting and prediction.

Key words: engineering of communications and transportation, free traffic, congested traffic, jam traffic, G-P algorithm, phase space reconstruction