Network assignment problems in transportation systems is a process that requires substantially high computational effort to complete a traffic flow analysis. This is due to finding of shortest path that connects every origin-destination pair in the network is to be determined prior to assignment process. This research is aimed at elaborating a new methodology of such shortest path finding by using the capability of taxonomy systems and genetic algorithms. Combination of the two developed in this research is called Genetic Taxonomy Evaluator (GTE) which is expected to be an effective and efficient tool compared to conventional method that enumerates any possible combination of links within the network. Such enumeration is explosive exponential in nature, so appears to be difficult to deal with practical problems in which networks may consist of thousands of links and nodes. GTE is a hybrid system of Taxonomy Re-constructor (TR) and Genetic Algorithm (GA) which are techniques originally developed in the discipline of computer learning of Artificial Intelligence. While keeping the characteristics of transportation networks TR transforms the networks into taxonomy structure which is hierarchically shaped based on problem to be solved. In the process TR also creates classification of nodes in the network. This classification provides facilities to isolate the problem to the core, and criteria that can be inserted in the genetic algorithm. Furthermore, the methodology is to be analyzed to comprehend the performance and benefits compared to conventional method. So a package program for GTE is developed in C Language. Results of analysis show that several optimal parameters should be determined prior to searching the shortest path, and the findings are quite promising as a threshold for further researches. (A)
Samenvatting