Abstract. This paper presents an improved Genetic Algorithm to solve the Transportation Network Design Problem (CTNDP) with interactions among different links. The CTNDP is formulated in an optimal design as a bi-level programming model. A key factor in the present approach is the combination of diploid based complex-encoding with meiosis specific features. The novel mutation operator proposed is another improvement that leads to a better robustness and convergence stability.The computational results obtained by comparing the performance of the proposed algorithm and other Genetic Algorithms for a test network demonstrates its better local searching ability, as well as its high efficiency.Finally, suggestions for further research and extensions are given.Key words: Genetic Algorithm, bi-level programming, Network Design Problem, complex-encoding.
The Transportation Network Design Problem -general descriptionThe Transportation Network Design Problem (TNDP) involves optimal decisions in determining a set of design parameters for improving an existing transportation network in response to an increasing level of traffic demand. The general increase in flow level results in traffic congestion, delays, higher fuel and maintenance costs, air pollution and accidents. The improvements of a transportation network, such as expansion of the capacities of the existing congested links, addition or deletion of links, traffic signal control adjustment, are made in accordance with a system optimum while considering the travel and route choice behavior of network users. The system optimum usually represents the minimization of the total travel time and construction costs.The network user's decisions correspond to a set of nonlinear relations that are formulated as an independent mathematical programming problem.In fact, a transportation network improvement involves the interaction of two parties with own objectives: the network planner represented by the transportation system authority and the network users that use the provided services. The traffic authority tries to optimize some overall objectives in the network, while the network users try to minimize their travel times/costs or perceived travel times/costs.The decision variables of the network planner affect the route choice behavior of network users which is based on two equilibrium principles:-the deterministic user equilibrium (DUE) condition [1] wherein network users choose the route with the shortest travel time/ the lower travel cost and equilibrium is reached where no user can unilaterally change routes to improve his/her own travel time or cost; assumptions of DUE can be somewhat unrealistic because of the variations in network conditions, variations in demand and no perfect information available for network users; -the stochastic user equilibrium (SUE) condition [2] where no user can unilaterally change routes to improve his/her own perceived travel time or cost; SUE may reflect travelers' behavior more realistically than DUE. One can consider that DUE is a special...