A new version of k-random walks algorithm in peer-to-peer networks utilizing learning automata

One of the most important issues in peer-to-peer networks is locating objects among a lot of data. There have been different search methods to find data with certain advantages and disadvantages. In this paper, we propose a new version of k-random walk algorithm utilizing learning automata. In this distributed method, the value of k for k-random walk is not selected randomly but it is selected in an adaptive manner. It is decided which walker is more useful to be selected in order to keep on the search according to past experience of each node. Simulation results show that the novel search algorithm improves the number of hits per query, success rate, generated messages per query and objects discovery delay in comparison with the k-random walk algorithm.