Traffic flow optimization bloat control genetic programming algorithm
In addressing the issue of increasing individual average size with iterations of the GP algorithm for training hyper-heuristic strategies in dynamic traffic assignment,this paper proposed various methods for bloat control in GP to constrain the genetic inheritance of large-sized individuals within the population.This enabled the algorithm to discover smaller-sized yet higher-performing hyper-heuristic strategies during the training process.Considering the potential performance differences of hyper-heuristic strategies on various structured road networks such as grid,ring radial,and free style networks,this paper adopted different structured road networks to train hyper-heuristic strategies for analysis and comparison.The trained hyper-heuristic strategies underwent simulation testing on urban road networks of varying scales and traffic volumes.The conclusion is that the double tournament bloat control method performs the best for different structured road networks.The GP algorithm obtained from this study demonstrates its ability to achieve shorter overall average travel times for road networks compared to existing dispatch methods.Consequently,more concise and effective hyper-heuristic strategies are derived,leading to en-hanced decision-making efficiency.
万方数据
维普
