Traveler's Route Choice Modeling And Its Impact Analysis Of Stochastic Traffic Network

Stochasticity is the key characteristic of the real-world traffic network.In the context of the stochastic traffic network,this dissertation analyzes traveler's route choice behavior,and proposes risk-averse traveler's route choice models based on the risk measure theory,behavioral economics theory and bounded rationality theory.Furthermore,the impact of traveler's route choice behavior on the traffic network is analyzed from the perspective of traffic network equilibrium and the total system travel time reliability.The study in this dissertation can show some implications for the design and implementation of the traffic policies,and for the real-time traffic management and control.More details are listed in the following.Firstly,this dissertation proposes the non-expected route travel time(NERTT)model,which belongs to the rank-dependent expected utility theory.The NERTT model is proved to be risk-averse under certain conditions.The NERTT can be reduced to the travel time budget(TTB)model and mean-excess travel time(METT)model with two different methods,which are based on the specific distortion function and based on a general distortion function,i.e.,a general framework for the risk-averse traveler's route choice is proposed.The NERTT is further extended to the non-expected risk-averse user equilibrium model,which is formulated as a variational inequality(VI)problem.The VI problem is solved with the proposed heuristic gradient projection algorithm with column generation.The numerical tests are conducted on the hypothetical traffic network to demonstrate the relationship between traveler's route choice and equilibrium flow pattern,and the numerical tests are also conducted on some large-scale traffic network to demonstrate the effectiveness of the proposed solution algorithm.Secondly,this dissertation proposes the subjective-utility route travel time(SURTT)model.In the TTB model and the METT model,a predefined confidence level needs to be given.However,due to the day-to-day travel time variations,the travelers cannot specify a proper confidence level.In the SURTT model,the confidence level is assumed to belong to an interval,and three different utility functions are adopted to weigh each confidence level in the interval,which show traveler's behavioral insights.The relationship between the SURTT model and the TTB,the METT and the NERTT model is discussed.The SURTT is further extended to the subjective-utility risk-averse user equilibrium model,which is formulated as a VI problem.The VI problem is solved with the proposed heuristic method of successive averages algorithm.The numerical tests are conducted on the hypothetical traffic network to demonstrate the relationship between traveler's route choice and equilibrium flow pattern,and the numerical tests are also conducted on the multiple-OD traffic network to demonstrate the effectiveness of the proposed solution algorithm.Thirdly,this dissertation proposes multiple criteria target-oriented route utility model.Motivated by Simon's satisficing model and considering three different influencing criteria,i.e.,route travel time,route late arrival penalty and route distance,the risk-averse traveler's multiple criteria target-oriented route utility model is proposed.The dependence between different route choice criteria is considered,which is captured by the copula.The exact form of the copula is found via the proved comonotonicity.The interaction among the targets is also considered,i.e.,the complementarity and the substitutability,which shows traveler's behavioral insights.The multiple criteria target-oriented route utility model is further extended to the multiple criteria target-oriented user equilibrium model,which is formulated as a VI problem.The VI problem is solved with the proposed heuristic method of successive averages algorithm.The numerical tests are conducted on the hypothetical traffic network to demonstrate the impact of the target on the equilibrium flow pattern,and the numerical tests are also conducted on the multiple-OD traffic network to demonstrate the effectiveness of the proposed solution algorithm.Finally,this dissertation proposes the moment-based method to evaluate total system travel time(TSTT)reliability,which shows the systematic performance of the route choice behavior.The unreliability function is proposed to reflect the TSTT reliability,and the upper bound of the unreliability function is calculated to show the reliability.Based on the moment of the uncertainty in the traffic network,the generalized moment problem(GMP)is conducted to calculate the upper bound,which belongs to the distribution-free method.The solution algorithm for the GMP consists of two components,which are the de-indicator and Lasserre's relaxation,which reformulates the GMP into a semi-definite program.Based on the first-order and second-order moment,the tightness of the upper bound is investigated,and based on the higher-order moment,the effectiveness of the higher-order information is demonstrated.