This paper proposed a global optimization method for solving the combined traffic assignment and signal control problem. Conventionally, the problem is formulated into a bi-level program: the upper-level describes the constraints of signal variables and minimize the performance index of the system, whereas the lower-level represents the equilibrium flow pattern. It is well known that bi-level program is non-linear and non-convex and it is hard to find the global optimization solution. In this article, we present a global optimal method to solve this problem. Firstly, we transform the bi-level program into a single-level program. Then, we implement an efficient mixed-integer model referred as DLog (Logarithmic disaggregated convex combination model) to linearize and approximate the non-linear constraints by piece-wise linear functions. Finally, the bi-level program is cast into a mixed-integer linear model, which can be easily solved by commercial mixed-integer program solver and thus the global optimal solution can be guaranteed. In numerical studies, the application of the method are demonstrated by a small network and compared with the results reported in the literature.
↧