On the Implementation of High Performance Computing Extensionfor Day-to-Day Traffic Assignment

The ability to find near-optimal traffic assignments for a very large network within a few days time will substantially contribute to reduce economic losses following major earthquake disasters. However, lack of efficient numerical tools and methodologies to solve this NP-hard problem within a reasonably short time is a major challenge. This paper presents details of an HPC (High Performance Computing) enhanced system which is developed to address this need of finding near-optimal traffic assignment for large networks in a short time. The developed system is based on day-to-day algorithm and enhanced with a distributed memory parallel extension to accelerate the computation by utilizing computer clusters or supercomputers. Details of basic implementations, strategies to accelerate the serial computations and parallel scalability, and numerical examples to demonstrate the effectiveness of the proposed strategies are presented.