The concept of Learning Automata (LA) operating in an unknown environment was initially introduced by M. L. Tsetlin. He studied the learning behaviors of finite deterministic automata under stationary random environment R (C1, C2, …,Cr) and showed that they are asymptotically optimal under some conditions. Following his pioneering work, Varshavskii and Vorontsova [2] found that stochastic automata also have learning performances. Since then, the learning behaviors of stochastic automata have been studied quite extensively by many researchers. The LA theory has now reached a relatively high level of maturity, and various successful applications utilizing learning automata have so far been reported. Despite the current matured state concerning the theory of LA and its applications, there are still several problems to be settled. Two of the most important are the insufficient tracking ability to the nonstationary environment and the relatively “slow” speed of learning. In my talk, I will briefly touch upon the literature on the learning behaviors of LA under nonstationary environments. Further, I will also briefly touch upon the learning mechanism of the hierarchical structure learning automata (HSLA) having been developed for solving the second problem. Following these introductory remarks, I will introduce our recent studies concerning the application of learning performance of the HSLA under nonstationary multiteacher environment. I will also introduce our recent studies which apply HSLA to some of the commercial games.