A Verification Framework for Fictitious Play Based Learning Algorithms

Distributed optimisation techniques have gained increasing attention due to fast development of autonomous robots. Many algorithms have been proposed to make optimisation more efficient. In this paper we propose a framework, which is based on probabilistic verification techniques, in order to compare the performance of various game-theoretic algorithms, in particular, fictitious play and its variants, after a finite number of iterations. To demonstrate the effectiveness of the framework, we apply the framework to a game which is inspired by wireless communication network problems, on five variations of fictitious play algorithms.