Learning to control dynamical systems

The rapid changes and increasing complexity of social and technological systems had lead to the demand of more powerful control theories and systems. Although several adaptive control techniques have been introduced, it is not strong to many real-world problem domains where the degree of uncertainty is too high and the computational complexity grows geometrically with the number of unknown parameters and its inherent linear nature of an assumed dynamic model. With an environment which is only partially known, it is desirable that the system be able to gradually learn the characteristics of the environment so as to improve its control strategy. Hence, there is a need for systems which possess some form of “intelligence” to be robust in nature. The system should be able to learn from its experience in order to achieve some form of “intelligence”. This paper describes a few works done on applying machine learning techniques to control problems