Parameter Optimization And Structure Automation Design Of Human-Simulated Intelligent Controller

Evolutionary algorithm is a random-searching method derived from biological evolution. It uses simple encoding technology to describe all kinds of complicated structures. Through a set of encoding representation it makes simple evolutionary operation and natural selection mechanism of the survival of the fittest, which instructs learning and determines searching direction. It is simple but robust. Especially it cannot be restricted by the searching space and does not need auxillary information, so it has been used in greatly broad fields and got many productions.Through the advantage in solving diffcult control problems, it shows Human-Simulated Intelligent Control (HSIC) theory based on Sensor-motor Intelligent Schema (SMIS) provides an effective method for complicated system problems. But during this process, human-simulated intelligent controller still has the problems of multi-parameters optimization and structure automation design. On the basis of the comparability with evolutionary thought and human-simualted intelligent controller, this dissertation invites the evolutionary thought to solve problems of HSIC.In the auto-control domain, the pendulum system is a kind of typical complicated control object which is nonlinear, multi-variable, under-actuated. It is not only the touchstone of verifying a new control theory, but also a basic platform of all kinds of control experiments. So through the results of simulation experiments, the dissertation is proved that the evolutionary algorithm can solve the problems of the double inverted pendulum based on HSIC.The main research contents of the dissertation are as follows:(1) Developing the improved genetic algorithm (IGA) and the improved genetic programming (IGP). The dissertation, in detail, describes all kinds of improved methods for simple genetic algorithm (SGA), which provides the methodological and technological support for the latter part of the dissertation.(2) Applying IGA to realize model parameters and control parameters optimization of double inverted pendulum. The dissertation specifically clarifies how to solve this above problems, and gives the results of simulation experiments. It is proved the validity of this aboved method.(3) Applying IGP to solve the structure automation design of double inverted pendulum. This dissertation describes the ideas of the controller structure automation design integrated the HSIC theory based on SMIC. It gives its design process and experiment results through a simple simulation experiment. It is showed the validity of IGP method.Evolutionary algorithm is effective in solving double inverted pendulum based on HSIC, which gives a methodological support for HSIC theory, and also extends the application of HSIC.