Delivery Route Optimization with Neuron Network

The method of transportation route optimizing is developed. It consists in estimating the time required to perform the transportation task using different routes and choosing a route providing the least required time. To estimate required time neural network is used. It can be trained using real information or simulation results. The method for obtaining the data for neural network training has been developed. The method is based on the simulation model of the process and allows simulating a transportation process in different conditions. With this model data for network training was received. Numerical experiments were conducted. The structure of the network was determined. This structure includes two layers with hyperbolic tangential activation function at the input and linear one at the output. Estimation of training data set minimal representative volume was fulfilled. The developed optimization method was checked. Variants of transportation on different days of the week, time of day and season are considered. 32 different variants of transportation process were considered.