An Analysis Of The Stochastic Traffic Assignment Paradox Of Two Combinational Route Choice Model

Route choice model is an important research topic in traffic assignment,which plays an important role in understanding travelers' travel behavior and enhancing the precision of traffic prediction.Recent years,a multiplicative hybrid model and a linear combinational hybrid model have been proposed on the basis of the Multinomial Logit and the Multinomial Weibit models,the two combinational models have better properties.Under the non-congestion assumption,this paper investigates the conditions for generating the stochastic traffic paradox for the two combinational route choice models,and studies the relationship between the paradox features of the four models,analyzes the possible discrepancies when applying these models to predict the traffic paradox under practical contexts.Firstly,this paper studies the stochastic traffic paradox that marginally improve the cost of a road in a network yet increases the total travel cost of the network.On a two-route network,this paper studies the conditions for generating stochastic traffic paradoxes when using the two combinational route choice model in stochastic traffic assignment,it is found that the conditions for the paradox to happen are closely related to the conditions for generating the paradox under the Multinomial Logit and the Multinomial Weibit model.Based on this,the paper further studies the stochastic traffic assignment paradox of the two combinational route choice model on the single O-D network with multiple independent routes and the single O-D network with overlapping links,the paradox features of the two combinational models in these networks show consistency with the results in the two-route network.When marginally improve a link in a multi-OD network,the paradox of each O-D pair does not affect each other,whether the paradox exists in the total network is determined by the sum of the link's marginal cost in each O-D pair.The paper also studies the discrepancies when applying these models to predict the traffic paradox under practical context.A simulation method based on Probit model is applied to generate valid route set of certain O-D pairs in the Sioux-Falls network,route samples are drawn from the route set.The maximum likelihood method was used to estimate the parameters of the four route choice models based on the generated samples,and the relationship between the estimated parameters and the goodness of fit are studied.Based on that,the paper studies the differences when predicting the stochastic traffic paradox of each road under different models.The paper relatively comprehensively studies the features of stochastic traffic paradox of the two combinational route choice models,and uses simulation data to corroborate that inconsistencies may occur when using different models to predict traffic paradox.The study enriches the theoretical foundation of the two combinational route choice models,deepens the understanding of the two models,promotes the application of the model in practice,and provides theoretical guidance for road network design and planning.