A stochastic optimization framework for road traffic controls based on evolutionary algorithms and traffic simulation
Junchen Jin,Xiaoliang Ma,Iisakki Kosonen
DOI: https://doi.org/10.1016/j.advengsoft.2017.08.005
IF: 4.255
2017-12-01
Advances in Engineering Software
Abstract:Traffic flow is considered as a stochastic process in road traffic modeling. Computer simulation is a widely used tool to represent traffic system in engineering applications. The increased traffic congestion in urban areas and their impacts require more efficient controls and management. While the effectiveness of control schemes highly depends on accurate traffic model and appropriate control settings, optimization techniques play a central role for determining the control parameters in traffic planning and management applications. However, there is still a lack of research effort on the scientific computing framework for optimizing traffic control and operations and facilitating real planning and management applications. To this end, the present study proposes a model-based optimization framework to integrate essential components for solving road traffic control problems in general. In particular, the framework is based on traffic simulation models, while the solution needs extensive computation during the engineering optimization process. In this work, an advanced genetic algorithm, extended by an external archive for storing globally elite genes, governs the computing framework, and in application it is further enhanced by a sampling approach for initial population and utilizations of adaptive crossover and mutation probabilities. The final algorithm shows superior performance than the ordinary genetic algorithm because of the reduced number of fitness function evaluations in engineering applications. To evaluate the optimization algorithm and validate the whole software framework, this paper illustrates a detailed application for optimization of traffic light controls. The study optimizes a simple road network of two intersections in Stockholm to demonstrate the model-based optimization processes as well as to evaluate the presented algorithm and software performance.
computer science, interdisciplinary applications,engineering, multidisciplinary, software engineering