System Identification and Control using Probabilistic Incremental Program Evolution Algorithm

An indispensable ability for intelligent control is to comprehend and learn about plants, disturbances, environment, and operating conditions. In this paper, the Probabilistic Incremental Program Evolution (PIPE) algorithm, with its self-organizing and learning ability, is used as a promising tool for such purposes. The previous work on evolutionary control by using tree structure based evolutionary algorithm was inverse control in general. In this case, Genetic Programming (GP) was usually used to evolving a directly control law of nonlinear systems. It is difficult to design a better fitness function that should reflect the characteristics of nonlinear systems, and a prior knowledge about operating conditions is usually needed. In this paper, a new identification and control method is proposed without prior knowledge of the plant. Firstly, the input-output behavior of the discrete-time nonlinear system is approximated by the individual structure of PIPE (PIPE Emulator). Secondly, a model based evolutionary controller (PIPE Emulator-based controller) of nonlinear system is designed. Simulation results for a typical nonlinear discrete-time system show the feasibility and effectiveness of the proposed method.