event

PhD Defense by Philip K. Pecher

Primary tabs

Title: A DDDAS Framework for Managing Online Transportation Systems


Philip K. Pecher
Ph.D. Candidate
School of Computational Science and Engineering, College of Computing

Industrial and Systems Engineering, H. Milton Stewart School of Industrial and Systems Engineering
Georgia Institute of Technology

Date: Monday, March, 26th, 2018
Time: 12 PM – 1:45 PM (ET)
Location: KACB 2100 (research-wing conference room)


Committee:
------------------------
Prof. Richard M. Fujimoto (Advisor, School of Computational Science and
Engineering)

 

Prof. David Goldsman (School of Industrial and Systems Engineering)

Prof. Michael P. Hunter (School of Civil and Environmental Engineering)

Prof. Michael O. Rodgers (School of Civil and Environmental Engineering)


Dr. Brian Swenson (School of Electrical and Computer Engineering, Georgia Tech Research Institute)



Abstract
------------------------

Dynamic Data-Driven Application Systems (DDDAS) integrate an executing simulation with data instrumentation in a feedback control loop.  Additional data can be assimilated into the application computations and the application may control the data acquisition process. Thus, the accuracy of the model may be improved or the simulation execution accelerated. DDDAS has been used in a variety of domains, including monitoring the spread of wildland fires, emergency & disaster management, construction & waste management, monitoring weather conditions & ocean forecasting, optimization of supply networks, and improving healthcare operations.

 

This thesis discusses a DDDAS for monitoring transportation systems by integrating trajectory prediction and accelerated microscopic traffic simulation. We first discuss a set of trajectory prediction methods from which likely trajectories can be sampled from efficiently. Afterwards, we show how microscopic traffic simulations can be driven by these estimates and computationally accelerated with a lazy evaluation and speculative execution scheme, termed Superimposed Execution. This technique can be used to simulate objects that travel in a spatial network and where collected output statistics are limited to a single entity. Under mild assumptions, this novel scheme yields the same results as explicitly enumerated runs, while - in some cases - approaching a speedup equal to the number of runs.

 

Lastly, a general computational method - termed Granular Cloning - is proposed to accelerate ensemble studies. Many runs of a computer simulation are needed to model uncertainty and evaluate alternate design choices. Such an ensemble of runs often contains many commonalities among the different individual runs. Simulation cloning is a technique that capitalizes on this fact to reduce the amount of computation required by the ensemble. Granular Cloning is proposed that allows the sharing of state and computations at the scale of simulation objects as small as individual variables, offering savings in computation and memory, increased parallelism and improved tractability of sample path patterns across multiple runs. The ensemble produces results that are identical to separately executed runs. Whenever simulation objects interact, granular cloning will resolve their association to subsets of runs though binary operations on tags. State sharing occurs at any scale allowed by the computer system and only the state that changes from the shared state is physically cloned.

 

Status

  • Workflow Status:Published
  • Created By:Tatianna Richardson
  • Created:03/12/2018
  • Modified By:Tatianna Richardson
  • Modified:03/12/2018

Categories

Keywords