Efficient and scalable multi-geography route planning
Proceedings of the 13th International Conference on Extending Database Technology
| Hi-index | 0.00 |
Simulators have been widely used for recreating conditions for the purpose of testing in different domains. Emerging applications require simulators to support large scale and complex activities, and to support testing and training at multiple levels of abstractions. Modeling such large scale scenarios requires participation of multiple simulators operating toward a common goal, and scalability is a fundamental issue. Modeling such scenarios results in large volumes of queries and updates to the system and is impacted by computationally intensive planning algorithms across large geographical areas. The focus of this thesis is to address scalability challenges across multiple levels in large scale activity simulations. Using evacuation as a driving activity, we will address scalability in both integrated simulations and single simulators. The first part of the thesis addresses scalability in a micro simulator that models an activity in micro time slices and on high resolution geographical spaces. In systems where agents drive the activity by programmed behavior, scalability issues become pronounced with the presence of large number of agents, and complex resource intensive operations like path planning. In this thesis we develop an efficient path reuse strategy which aids participating agents in quickly computing evacuation routes to common safe destinations. The techniques proposed in this thesis have been evaluated in the context of an agent based activity simulator (DrillSIM) that we have designed. The second part of the thesis deals with scalability at the integrated simulation level. Here activity planning is affected by the presence of multiple simulators which operate on heterogeneous data sources. Specifically, the activities must span several geographical maps, and the heterogeneity and scale of participating geographies contribute to scalability challenges. This thesis will address the multi-geography route planning problem in the context of a large scale evacuation, where different simulators both at micro and macro level participate. The proposed solution involves an efficient algorithm for planning paths across different geographies along with precomputation strategies to speed up planning. We study our strategies in the MetaSIM platform, a web-based platform for the integration of simulation tools developed for disaster response, operating on a campus geography.