Structured Modeling in Large-Scale Dynamic Transportation System Optimization: Microscopic State-Space-Time Network and Macroscopic Computational Graph Based Framework
Xuesong Zhou
Ira A. Fulton Schools of Engineering, Arizona State University
ABSTRACT: Zhou presents two new modeling frameworks, a microscopic state-space-time (SST) network and a macroscopic computational graph (CG), for two major applications: to optimize vehicular routing decisions with pickup and delivery time windows (VRPPDTW), and to fully incorporate multiple data sources, including loop detector counts, and GPS location samples, for jointly estimating traffic congestion dynamics and driving behavior parameters. These two modeling frameworks enable us to prebuild many complex state transition constraints and relationships into a well-structured hyper network, so that the resulting optimization model can be nicely reformulated as multi-commodity network flow models with a very limited number of side constraints. The resulting relaxed problems can be solved by computationally efficient dynamic programming algorithms within a Lagrangian decomposition or a backpropagation based nonlinear optimization framework.
