Research On Route Choice Behavior Under Risk Based On Prospect Theory

Route choice analysis plays an important role in the prediction of urban traffic state,design of effective traffic control and guidance strategies,and implementation and optimization of transportation infrastructure layout designs,etc.Due to stochasticities in both transportation demand and supply and also the temporal and spatial interactions between them,urban transportation network is full of uncertainty.Travelers behave boundedly rationally in their route choices under risk and uncertainty.This paper tries to model travelers' bounded rationality in route choice decisions from three aspects,i.e.,perception of network state,evaluation of route prospect(utility)and route choice making based on the related evaluations,which partly form the basic decision hierarchies for route choices.We described travelers' boundedly rational behaviors in those decision hierarchies,characterized those behaviors with proper theoretical models,and developed user equilibrium models based on those behavioral models.Then,we provided numerical examples to demonstrate the flexibility and capability of the proposed models in capturing travelers' bounded rationality in route choice decision,and investigated the effects of those boundedly rational behaviors on the equilibrium flow distribution and associated network performances.This paper comprises the following several topics:(1)Research on the impacts of stochastic perception error(SPE)on individuals' perception of route travel time and thus their route choices.We investigated the effects of SPE on perceived travel time distributions,and used the latter to substitute the probability weighting function in prospect theory model and developed a prospect theory model based on perceived travel time distribution.Perceived travel time distribution relaxes the assumption on route travel time distribution patterns which are often hard to obtain in urban transportation networks,and ensures the properties of probability distribution at the same time(in comparison,the probability weighting distribution may not ensure the sum of cumulative probability weighting to equal 1).Numerical results show that,the two SPE parameters(mean and variance)affect the perceived travel time distribution in different ways,the SPE mean parameter affects both the expectation and variance of perceived travel time distribution,while the SPE variance parameter only impacts on the variance of perceived travel time distribution.Under perceived travel time distribution,the equilibrium flow distribution will be affected,and the network total expected travel time and average link Volume/Capacity ratio fluctuate in opposite directions.(2)Research on travelers' boundedly rational behavior in the route choice making hierarchy.We assume that travelers adopt the rank-dependent decision scheme(RDDS)to make route choice decisions,and they choose among routes that satisfy their rank-preference threshold.We analyzed the flexibility of RDDS in modeling varied rationalities in route choice making and in controlling the size of individual and population potential route set.Then,we developed a bi-criteria rank-dependent user equilibrium model in a congested network based on the RDDS,which included a continuously distributed value of time(VOT)parameter to describe the heterogeneous preference between route travel time and the fixed toll.The new model provides a potential alternative to the boundedly rational user equilibrium model based on “threshold-indifference band” approach.(3)Combine the theoretical models in(1)and(2)and develop a complex route choice behavior model under risk.We proposed two reference point models with multi-criteria consideration,including the “ideal reference point” which takes the best value among all for each criterion and the “route based reference point” which takes the route with minimal generalized disutility.Numerical results showed that travelers' bounded rationality in route choice making might improve the network performances in both total travel time and toll;while assuming ideal reference points may not lead to improvement of network performances.Besides,this paper provided different solution algorithms for each user equilibrium model.Numerical examples demonstrated the effectiveness of the proposed algorithms.The paper concluded with discussion of some future research directions.