Summary

The California Department of Transportation has recently embarked on an ambitious state-wide program to support the use of microscopic traffic simulation models. A recent survey shows that in California alone there are some 80 on-going or recently completed projects (mostly sponsored by Caltrans) use traffic simulation to study a wide range of problems, including interchange analysis, corridor traffic management, reconstruction planning, air quality analysis and growth impact assessment. Among these eighty projects, more than 60% use high fidelity traffic simulation models such as Paramics. A case could be made that Caltrans is at the very forefront of any state agency, perhaps in the whole world, in encouraging the use of microscopic simulation as a viable option for traffic analysis. That there were several in-house presentations on detailed use of microscopic simulation at the annual Caltrans Traffic Engineers Conference in Sacramento in August 2002, is evidence of this. While demonstrating the significant benefits from using microscopic simulations in detailed analysis for operational improvements and even highway design alternatives, the discussions in the conference also showed the primary deficiencies that need immediate attention. Perhaps the most important can be identified as,

• Difficulties in handling large networks comprehensively and building the simulation capabilities incrementally, with proper attention to the speed and data-set issues.
• Difficulties in handling the development of contextually accurate simulation models, primarily focusing on the calibration and fine-tuning issues.

• Type of distributed processing techniques to be used
• The key traffic issues involved, primarily the interface traffic flow between sub-networks among which the simulation effort is distributed, and the O-D patterns and paths of vehicles in each sub-area that involve other sub-areas as well.
• The computational issues involved in the platform used for distribution
• he issues in streamlining the data preparation effort when sub-area data sets are developed seperately (the topic of network "stitching")
• he database management issues when multiple network data sets are involved

1. Survey the current district users to identify key limitations encountered with respect to size and speed in microscopic simulations.
2. Survey the current district users and Caltrans traffic analysis policy makers on the potential future plans on large-network simulation and data management, as well as plans for migration to or augmentation of analysis with simulators other than Paramics.
3. Synthesize the current state-of-the-art in large network modeling, specifically focusing on the distributed and parallel simulations. Any past use of Paramics as well as other commercial microscopic simulators for such large networks, will be examined.
4. Develop the test platform for distributed large network simulation.
5. Develop the methodological details on distributing the simulation and prepare the algorithmic schemes for different types of distribution (light master controller and heavy simulators vs heavy master controller and light simulators, as two extremes detailed below)
6. Develop methodological details of network stitching and techniques to handle data of different granularity and
7. Identify candidate study networks (large area network)
8. Prepare the data sets for study
9. Code the algorithms into micro-simulation APIs and develop the distributed framework.
10. Compare the benefits and disadvantages of the different types of distribution schemes from Task 5
11. Develop guidelines on large network decomposition and/or composition of the large network with multiple data sets, and on establishing large network modeling platforms at the Caltrans analysis offices.
12. Conduct workshops to transfer the project results to Caltrans local district users and any other associated users.