5390

Efficient Execution on GPUs of Field-Based Vehicular Mobility Models

Kalyan S. Perumalla
Oak Ridge National Laboratory, Oak Ridge, Tennessee, USA
22nd Workshop on Principles of Advanced and Distributed Simulation, 2008. PADS ’08

@inproceedings{perumalla2008efficient,

   title={Efficient execution on GPUs of field-based vehicular mobility models},

   author={Perumalla, K.S.},

   booktitle={Principles of Advanced and Distributed Simulation, 2008. PADS’08. 22nd Workshop on},

   pages={154–154},

   year={2008},

   organization={IEEE}

}

Download Download (PDF)   View View   Source Source   

540

views

Large-scale scenarios of vehicular traffic simulation problems are characterized by complex queuing effects, control mechanisms and other interactions of the traffic on the control and vice versa. While small-sized scenarios are relatively easy to explore and analyze, larger scenarios need specialized treatment for efficient execution. The simulation challenges of speed and scale become pronounced when network sizes are very large (millions of road intersections) and/or vehicular traffic load is immense (several million simultaneously active vehicles). Commonly used execution approaches roughly correspond to the two extremes of the simulation spectrum. Aggregate models correspond to one extreme, typically employed to achieve coarse results at a relatively high simulation speed. Micro-simulation represents the other extreme, employed for finer granularity at the expense of greatly decreased simulation speed. Hybrid methods are emerging as via media that provide medium level modeling granularity while still affording fast execution. In some of our active projects in real-time data- driven simulations, we are developing one such hybrid method. Two novel aspects of our hybrid method are: (1) formulation of a field-based model of vehicular traffic movement in a large road network (2) fast execution of the field-based hybrid model on data-parallel platforms of general-purpose graphics processing units (GPUs).
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2017 hgpu.org

All rights belong to the respective authors

Contact us: