Использование параметрически заданных устойчивых распределений вероятностей в задаче нахождения надежного кратчайшего пути

Abstract

The use of stable probability distributions in the reliable routing problem In this paper, we consider the reliable shortest path problem in a timedependent stochastic transportation network. The problem is to find a routing policy that maximizies the probability of arriving at the destination point on time. It is proposed to use parametrically defined stable probability distribution Levy to describe the travel time of road segments. The use of stable distributions allow us to replace the convolution operation with the distribution value, and significantly reduce the execution time of the algorithm. Experimental analysis have shown that the use of stable distributions allows to approximate the exact value of the arrival probability to a destination with a low approximation error.