Comparing Dynamic User Equilibrium and Noniterative Stochastic Route Choice in a Simulation-Based Dynamic Traffic Assignment Model: Practical Considerations for Large-Scale Networks

Table 1

A summary of literature on simulation-based traffic signal optimization studies.

Traffic assignment	Network size	Summary
Deterministic user equilibrium (DUE)	Local (<5 signals or <5 available routes)	Very few studies are found carrying out signal optimization using DUE in relatively small networks with few routing options including [66] that used DUE to calculate the reserve capacity of the network. Meanwhile, [67, 68] implemented DUE on a local network with isolated signal or few signals as well as with a real network of multiple signalized intersections.
	Global (>5 signals or >5 available routes)	DUE is the popular choice of traffic assignment for signal optimization in relatively large networks [67–70]. These studies yield efficient techniques for signal optimization based on DUE traffic assignment with some providing a comparison between DUE with SO, concluding significant advantage of DUE for adaptive signals.
Stochastic user equilibrium (SUE)	Local (<5 signals or <5 available routes)	SUE is more prominent in localized signal optimization [57, 68, 71, 72]. The study reported in [58] has provided a comparison of the performance of deterministic and stochastic user equilibrium techniques over different networks and observed improved results for SUE in local networks.
	Global (>5 signals or >5 available routes)	Very few studies are found in the literature using SUE in relatively large networks including [72] that has recommended the use of DUE on global traffic signal optimization.
System optimal (SO)	Local/global	References [67, 73, 74] used SO in both large and small networks. Reference [67] argued that DUE is advantageous over SO in designing adaptive signals in a network.
Stochastic route choice (SRC)	Local (<5 signals or <5 available routes)	Use of SRC mostly prevails in the studies on a single freeway or highway with few alternate routes available [65, 75–79].
	Global (>5 signals or >5 available routes)	A few studies are found in the literature that used relatively large networks and SRC [80–82]. In [82], the selection of route choice model was arbitrary as the main focus of the study on the optimization part of the problem rather than the simulation.