Network Design Using Chemical Reaction Optimization and Markov-Chain Traffic Assignment

摘要

Transportation network design and traffic flow prediction are essential tools in reducing transportation network congestion. In this paper, we propose a new approach to solving the transportation Network Design Problem (NDP). Several variations of this problem are investigated in literature, however, we study here an NDP where the decision is to select a set of roads for traffic direction conversion from two-way to one-way traffic. This traffic direction change allows us to expand the affected roads' flow capacity at the expense of restricting flow in opposite directions. When done strategically, this approach can result in significant congestion reduction for the transportation network as a whole. The new approach builds on the recently developed Chemical Reaction Optimization (CRO) metaheuristic and leverages Markov chain traffic assignment to model road-users' reaction to network modifications. We propose a modified adaptive version of the CRO metaheuristic allowing it to more efficiently identify good solutions using traits from the found best solutions as the search progresses. We use the city of Abu Dhabi to test the approach and report on our results. We also compare the modified adaptive CRO results to those of Genetic Algorithm (G A) to demonstrate the new approach's potential. Compared to GA, we find the new approach to be more efficient in finding better solutions faster, however, it is also more sensitive to parameters setup.