Coordinate Descent based Decomposed Anticipatory Dynamic Traffic Control

Traffic networks inherently exhibit several degrees of decomposition, both hierarchically and geographically, and traffic control has been, in practice, severely influenced by this decomposition. Anticipatory Traffic Control literature (Allsop, 1974), in its own, deals with the fact that user behavior can be, at least to some degree, predicted, and that an “optimal” controller should take this behavioral response into account when determining the best strategy to apply. While historically rooted in the field of Static Assignment, in recent years researchers have been looking into translating Anticipatory Traffic Control in dynamic networks, aiming at exploiting the inherent advantages on real-life instances (Taale, 2008), (Taale and Hoogendoorn, 2012), (Ukkusuri et al., 2013).