The book is a revised version of Ph.D. thesis focused on decentralized control. In some applications, the classical approach of controlling a large multivariable system by central controller has become unattractive and sometimes unusable for either economic or reliability reasons. The decentralized approach to the control of multivariable systems is studied in this book. The advantage of the usage of classical, well-known and verified control algorithms designed for one- dimensional systems was utilized. These algorithms become suitable for decentralized control by adding the on-line identification part. The supervisory logic can be used to increase the stability of on- line system model identification. The idea of logical supervisor is proposed to simplify the design of supervisory block while preserving the stability of identification process. The book also presents the Self-tuning Controllers Simulink Library, which was developed to give a framework for design and verification of self-tuning controllers. Usage of decentralized approach I verified by experiments performed on both simulation systems and real-time laboratory plants.