Learning automata-based algorithms for solving the stochastic shortest path routing problems in 5G wireless communication

Stochastic Shortest Path Routing (SSPR) problem has always been a critical topic in communication area, especially after a boom in the 5G wireless communication emerges. In this paper, we consider the SSPR problem in a network, where the context is a dynamic, and simultaneously the probability distribution of the constantly changing edges is unknown beforehand. A hierarchical structurized learning automata applied to SSPR, namely SSPR-hieraStructure LRI, is proposed to solve the issue, providing a novel way to pick out the shortest path from source to destination in a stochastic graph. Two kinds of LA prototypes, i.e, variable action learning automata and hierarchical system of learning automata are assembled to exert their respective strengths, based on which we further makes adequate use of a hierarchical structure for modeling. The innovation lies in a convergence by layer rather than as a whole, thus determining the shortest path is converted to searching an optimal node in each layer. Moreover, the searching space is largely reduced by structural pruning, for the purpose of speeding up the algorithm progress. The experimental results involving various aspects demonstrate the effectiveness and efficiency of the proposed algorithm, which outperforms the state of arts in LA based SSPR problems, with a higher accuracy rate, a faster convergence speed, a lower sampling consumption but a higher sampling efficiency.