A route choice optimization method for dynamic user equilibrium under strongly heterogeneous cost fields
Published:
The invention discloses a path optimization method for achieving dynamic user equilibrium under strong heterogeneous cost fields, which pertains to the field of traffic path distribution. This method does not require the pre-definition of precise grid structures and accurately determines the distance information of each location point by considering cost impacts. It integrates an improved NES strategy based on physical information, designed using a fully connected neural network, to determine the Eikonal decomposition factor. This factor, when multiplied by the distance information, results in the cost potential. The method reliably decomposes the Eikonal equation with high processing efficiency, facilitating the subsequent solution of the preset dynamic user equilibrium model to obtain the optimal initial path. Based on the new cost function at the i-th dynamic moment, the distance information from the previous moment, and the improved NES strategy, the cost potential at the i-th dynamic moment is obtained to determine the optimized path at that time. This approach efficiently solves the cost potential at the i-th dynamic moment through the concept of transfer learning.