Policy and Complex Systems
Figure 3 . The internal structure of the stochastic ABM showing attributes , behaviors , and the environment of four agent classes — state agency , RPCs , local towns , and projects .
with a genetic algorithm to minimize the difference between observed and simulated funding allocations . The refined baseline scenario presented in this study presents different parametric settings compared with what was reported in Zia and Koliba ( 2015 ). Furthermore , alternate scenarios tested in this study and questions addressed are markedly different from earlier study . Finally , the simulation horizon is extended to 50 years , compared with 25 years reported in earlier work .
Structurally in the ABM , as shown in Figure 3 , a state agent ( i . e ., VTrans ) contains 11 nested RPCs and 600 local towns that are nested within RPCs . Furthermore , transportation projects are modeled at the inner most layer of the nested hierarchy . RPCs , local towns , and transportation projects are thus modeled as multilevel nested agents in this ABM , whereby project class agents are spatially situated within local town agents , local town agents are situated
within RPC agents , and RPC agents exist within state agent . We have deliberately anonymized the identity of RPCs inside the ABM ; whereas local towns inside the RPCs are initialized on random distribution basis ( described below ).
Figure 4 shows a state chart and transition functions among different states for the agent class of transportation roadway projects . During 2010 focus groups , one of the experienced participants described V-TRANS decision heuristic for funding transportation projects as a “ funneling approach ” that is captured for modeling the state chart of the project class in the ABM model shown in Figure 4 . Every year , different agents in the policy implementation network identify a large number of transportation infrastructure problems . VTrans keeps an updated list of these problems and selects a sub-sample of these problems for undertaking feasibility
59