研究期間：10108~10207;This project is mainly concerned with transportation combined models which are to be studied in three consecutive years. In the first year, currently available supernetworks for different transportation combined models will be carefully reviewed and modified, if necessary, for better representation. Of course, supernetworks for new combined models will also be proposed. A supernetwork is usually defined as an augmented network that consists of a “basic network” for route choice and a “virtual network” for other travel choices. Supernetwork representations are useful pedagogical device to interpret various combined travel choice models as an extension of the fixed demand traffic assignment problem. A nested combined model consisting of four choices subject to variable demand will be taken for demonstration. Equilibrium conditions will be derived and solutions will be obtained by a suitable algorithm. In the second year, the focus will be put on the development of more efficient algorithms for the combined models. As pointed out by Sheffi (1985), multistage solution algorithms are more efficient than the FW method for solving combined travel choice models. However, in view of fast development of traffic assignment algorithms, the multistage solution algorithms are no longer good choices for solving combined models. Instead, recent quick-precision algorithms such as projected gradient, “B” or “OBA” algorithms could be more efficient by help of the proposed new supernetwork representations. Note that realizing a complex solution algorithm in depth as well as coding it correctly is always a formidable labor intensive work. In the third year, the combined travel choice models (using the supernetwork representations developed in the first year and computer code(s) for a combined model completed in the second year) will be further extended to dynamic scenario in which additional temporal dimension will be incorporated. Mathematical properties such as uniqueness and existence will be elaborated. Subsequently workable solution algorithms will be proposed and numerical results will be provided. Useful information for practical applications may thus lead the change of way of the modeling work by practitioners currently working for consulting firms.
