GPU-based parallel computing for activity-based travel demand models
conference paper
Activity-based travel demand models (ABMs) are gaining popularity in the field of traffic modeling because of their high level of detail compared to traditional travel demand models. Due to this, however, ABMs have high computational requirements, making ABMs hard to use for analysis and optimization purposes. We address this problem by relying on the concept of parallel computing using a computer’s graphics processing unit (GPU). To illustrate the potential of GPU computing for ABM, we present a pilot study in which we compare the observed computation time of an ABM GPU implementation that we built using NVIDIA’s CUDA framework with similar, non-parallel implementations. We conclude that speed-ups up to a factor 50 can be realized, enabling the use of ABMs both for fast analysis of scenarios and for optimization purposes. © 2019 The Authors. Published by Elsevier B.V.
TNO Identifier
869361
ISSN
18770509
Publisher
Elsevier B.V.
Source title
Procedia Computer Science, 10th International Conference on Ambient Systems, Networks and Technologies, ANT 2019 and The 2nd International Conference on Emerging Data and Industry 4.0, EDI40 2019, Affiliated Workshops, 29 April 2019 through 2 May 2019
Editor(s)
Shakshuki, E.
Pages
726-732