Y. Shi et al. (Eds.): ICCS 2007, Part I, LNCS 4487, pp. 1050 – 1057, 2007. © Springer-Verlag Berlin Heidelberg 2007 Ad Hoc Distributed Simulation of Surface Transportation Systems R.M. Fujimoto, R. Guensler, M. Hunter, K. Schwan, H.-K. Kim, B. Seshasayee, J. Sirichoke, and W. Suh Georgia Institute of Technology, Atlanta, GA 30332 USA {fujimoto@cc, randall.guensler@ce, michael.hunter@ce, schwan@cc}.gatech.edu Abstract. Current research in applying the Dynamic Data Driven Application Systems (DDDAS) concept to monitor and manage surface transportation systems in day-to-day and emergency scenarios is described. This work is focused in four, tightly coupled areas. First, a novel approach to predicting future system states termed ad hoc distributed simulations has been developed and is under investigation. Second, on-line simulation models that can incorporate real-time data and perform rollback operations for optimistic ad hoc distributed simulations are being developed and configured with data corresponding to the Atlanta metropolitan area. Third, research in the analysis of real-time data is being used to define approaches for transportation system data collection that can drive distributed on-line simulations. Finally, research in data dissemination approaches is examining effective means to distribute information in mobile distributed systems to support the ad hoc distributed simulation concept. Keywords: surface transportation systems, ad hoc distributed simulations, rollback operations. 1 Introduction The Vehicle-Infrastructure Integration (VII) initiative by government agencies and private companies is deploying a variety of roadside and mobile sensing platforms capable of collecting and transmitting transportation data [1-3]. With the ongoing deployment of vehicle and roadside sensor networks, transportation planners and engineers have the opportunity to explore new approaches to managing surface transportation systems, offering the potential to allow the creation of more robust, efficient transportation infrastructures than was possible previously. Effective and efficient system management will require real-time determinations as to which data should be monitored, and at what resolutions. Distributed simulations offer the ability to predict future system states for use in optimizing system behaviors both in day-to- day traffic conditions as well as in times of emergency, e.g., under evacuation scenarios. Data collection, data processing, data analysis, and simulations performed by system agents (sub-network monitoring systems, base stations, vehicles, etc.) will lessen communication bandwidth requirements and harness surplus computing