Experimental Study On The Influences Of Information On Travel Choice Behaviors

With the acceleration of urbanization in China,traffic congestion has become an important factor hindering the sustainable development of cities and affecting the quality of life of residents.To solve this problem,the traditional solution is to expand or improve the existing transportation facilities,but this method has been proved to be not perfect,and sometimes even counterproductive.In recent years,with the rapid development of information technology,many countries and regions have begun to build Advanced Traveler Information Systems(ATIS)to solve traffic issues.However,in the actual transportation system,traffic flow is the result of the macro-aggregation of various travel choices of travelers in limited time and space,whether ATIS can effectively regulate the travel choice behaviors of travelers and further optimize the spatial and temporal distribution of traffic flow is an urgent issue to be studied at this stage.Therefore,this paper studies the influences of traffic information on travelers' choice behavior through behavioral experiments.Firstly,this paper designs and carries out a series of real-life travel choice behavior experiments.There are two scenarios in the experiments:one is global information group,which can feedback all travelers' historical choices information;the other is local information group,which can feedback his/her own historical choice information.It is found that global information has no significant effect on the speed and quality of convergence to Nash Equilibrium,but it will accelerate the process of travelers'(especially female travelers')decision-making towards stability.The average income of travelers is significantly negatively correlated with the frequency of their own strategy changing,and in most cases,the low income of travelers leads to frequent strategy changing;Feedback information has a certain influence on the travelers' strategy response mode,and the travelers who receive the local information are more inclined to the positive response mode.Secondly,drawing lessons from some theories and methods in the field of human trajectory prediction,this paper analyses the travelers' decision sequences in the experiments.It is found that the decision processes of the travelers who receive global information feedback are more complex,but the mean potential predictability of the travelers' sequences is not significantly reduced.Using the n-order Markov chain-based predictors to predict the travelers' decision sequences,it is found that the 1-order model has the best prediction accuracy.Furthermore,considering the stationarity of the decision sequences,it is found that the mean prediction accuracy of the non-stationarity sequences in the global information group is significantly higher than that of the stationarity sequences,the mean prediction accuracy of the non-stationarity sequences in the local information group is not improved,and even the volatility is obviously getting larger.Finally,based on the heterogeneity of the travelers' behaviors,the experience weighted attraction(EWA)learning model is used to describe the travelers' decision-making processes.The corresponding parameters of each traveler are obtained by genetic algorithm,and the different preferences of travelers under different experimental scenarios are explained by the psychological meaning of parameters.From the training process and prediction results of the model,the EWA learning model can better describe the decision-making processes of travelers who receive global information feedback.