Comparing And Analyzing Intelligent Optimization Algorithms For Urban Transportation Continuous Network Design Problem

Designing effective algorithms for continuous network design problem is an important problem in transportation studies. In this thesis, four intelligent optimization methods for continuous network design problem are discussed based on existing studies. The works made in this study are summarized is as follows:(1) A bi-level programming model for continuous network design problem is introduced. The objective function at the upper level is defined as the sum of the total travel time on the network, and the total investment costs of link capacity expansions; the lower level problem is the user equilibrium assignment model. Four popular intelligent heuristic optimal methods, including the genetic algorithm, particle swarm optimization, ant colony algorithm and simulated annealing are used to solve the bi-level programming model. Numerical experiments show that the four methods can solve the continuous network design problem. However, the convergence level and implementation time of each method have clear difference under the different setting of the parameters and the algorithm structure.(2) The sensitivity analysis method is first time used to analyze and compare the influence of the different selection of parameters to the implementation of each intelligent optimization method. Suggestions of the selection of parameters are given. Numerical experiments demonstrate the efficiency and precision of these methods can be improved greatly with the proposed suggestions.(3) Chaos time series methods are introduced to analyze the complexity of the solution process of the genetic algorithm. Experiments demonstrate it is a feasible way to use chaos time series methods to analyze the complexity of the solution process of genetic algorithm. The largest lyapunov exponent can be used to manifest the complexity of genetic algorithm effectively. Comparing different examples, it is concluded that genetic algorithm become more complex with the increase of the scale of network.(4) The differences of the four intelligent heuristic optimal methods for the bi-level programming model of continuous network design problem are compared further from the qualitative and quantitative analysis. With the comprehensively consideration of indicators including convergence accuracy and implementation efficiency, it is concluded that particle swarm optimization is the best method to determine the optimal solution, genetic algorithm is less efficient. The ant colony algorithm can reach high convergence accuracy but computation time is more longer. The simulated annealing algorithm is the worst both from the comparison of convergence accuracy and the implementing time.