Calibrating Travel Demand in Large-scale Micro-simulation Models with Genetic Algorithms: A TRANSIMS Model Case Study

To better address emerging transportation issues, there has recently been an increased interest in the development and application of large-scale micro-simulation models for areas that traditionally had been modeled using a four-step model. Developing and calibrating these models, however, present several challenges, particularly with respect to calibrating the travel demand, which in this case, is dynamic and has to show how demand varies over a 24-hour period. This paper aims at providing a comprehensive assessment of the ability of Genetic Algorithms (GAs) to calibrate travel demand in large-scale micro-simulation models, utilizing both a prior known Origin-Destination matrices and field traffic counts. As a case study, the paper considers the Transportation Analysis and Simulation System (TRANSIMS) and uses two test networks: (1) a synthetic network, to first gain insight into the characteristics of the problem; and (2) a realistic network of the North Campus of the University at Buffalo (UB). Several ideas are proposed in order to: (1) speedup the run time for the TRANSIMS model; (2) reduce memory usage; and (3) reduce the search space of the problem. The GA is custom-designed for the problem, and is equipped with special operators and mechanisms to increase search efficiency. The results indicate that GAs appear to have a significant impact on improving the quality of the solutions. Specifically, for the UB campus network, running the GA for only 2000 evaluations resulted in a dramatic 76% reduction in the model’s Root Mean Square Error (RMSE), when comparing simulated and field counts.


  • English

Media Info

  • Media Type: DVD
  • Features: Figures; References;
  • Pagination: 16p
  • Monograph Title: TRB 89th Annual Meeting Compendium of Papers DVD

Subject/Index Terms

Filing Info

  • Accession Number: 01154634
  • Record Type: Publication
  • Report/Paper Numbers: 10-2437
  • Files: BTRIS, TRIS, TRB
  • Created Date: Jan 25 2010 11:09AM