Artificial Neural Network Delay Model for Traffic Assignment Incorporating Intersection Delay Costs
Document Type
Journal Article
Publication Date
2009
Subject Area
operations - traffic, infrastructure - traffic signals, ridership - commuting
Keywords
Travel time, TRANSYT-7F (Computer program), Traffic signal timing, Traffic signal settings, Traffic delay, Traffic assignment, Signalized intersections, Signalised intersections, Settings (Traffic signals), Root mean square error, Neural networks, Journey time, Artificial neural networks, ANNs (Artificial neural networks)
Abstract
Transportation planning models do not usually consider intersection delays explicitly in calculating path travel times. However, because intersection delays often make up a considerable portion of total travel time in urban areas, ignoring them will result in inaccuracy in model results. Incorporating intersection delays into a traffic assignment model presents several challenges. One is that future signal timing plans are unknown, and another is that introducing intersection delays into traffic assignment renders the necessary condition for convergence invalid. This paper proposes a pragmatic approach to considering intersection delays in traffic assignment by assuming that signal timing plans will be optimized for given traffic conditions and by demonstrating that a convergent solution can be obtained when intersection delays are explicitly accounted for. A large data set of traffic volume combinations representing the operation of an intersection and the corresponding intersection delays based on optimized signal timing are obtained from TRANSYT-7F. This data set is used to train an artificial neural network (ANN) model that provides delay estimates for given traffic conditions for certain types of intersection configurations. An iterative process of traffic assignment and delay estimation is carried out to obtain a stable assignment solution. The delay estimates from the ANN delay model are evaluated by the percentage root mean squared errors, which are less than 25.6%, with larger prediction errors typically associated with severely oversaturated conditions. An example is presented to show the use of the delay model in traffic assignment. The delay model allows the concurrent optimization of signal controls and traffic routing to obtain a local optimum network solution.
Recommended Citation
Ding, Zhen, Zhao, Fang, Wu, Yongqiang, (2009.) Artificial Neural Network Delay Model for Traffic Assignment Incorporating Intersection Delay Costs. Transportation Research Record: Journal of the Transportation Research Board, 2132, pp 25-32.
