Intelligent Control Of Triple Inverted Pendulum Based On CGA-PIDNN

To solve control problem of the inverted pendulum which has stable multivariable,nonlinear, dynamic instability and the two outstanding problems of traditional geneticalgorithm (GA, Genetic Algorithm)-low convergence rate and easily be dropped in localminimization, a method based on the dynamic characteristics of PID neural network(PIDNN, Proportional-Integral-Derivative Neural Network) optimized by geneticalgorithms cloud (CGA, Cloud Genetic Algorithm) is proposed to establish the nonlinearmodel of three inverted pendulum, CGA-PIDNN control system of three invertedpendulum is designed based on this nonlinear model, and compares it with the controleffects of PIDNN control system optimized by standard GA.Firstly, PIDNN is used to achieve three inverted pendulum modeling and controltasks. Various parameters of PIDNN have clear physical meaning, the structure is simple,and PIDNN has learning and memory function and nonlinear mapping capabilities ofneural network. PIDNN’s structures and algorithms are introduced. Aiming at the lack ofGA, CGA is used to optimize PIDNN and to establish nonlinear model of three invertedpendulum based on CGA-PIDNN. Hidden layer of PIDNN incorporates neurons that haveproportional, integral and derivative function, which makes PIDNN has a better ability ofdeal with dynamic information. Cloud droplets of normal cloud model have a stabletendency and randomness, while GA has a good parallel design structure and globaloptimization features. CGA combines advantages of normal cloud model and GA toproduct adaptive crossover and mutation probability by using normal cloud generator. Thismethod can increase the speed of convergence, the effect of network modeling and controlalso is improved.Secondly, CGA-PIDNN complete control system of triple inverted pendulum isestablished based on CGA-PIDNN nonlinear model of triple inverted pendulum.Compared with the control effect GA-PIDNN, indicating that CGA-PIDNN requires lessevolution algebra can converge, it has better the convergence speed than GA-PIDNN. Andto some extent it overcomes the problem less easy to fall into local minimum of GA,control effects are better. This is a new kind of methods nonlinear dynamic system modeling and control.Finally, in order to achieve observe the control process and trolley pole triple invertedSwing by multi-directional and of multi-angle, in LabVIEW (Laboratory Virtualinstrument Engineering Workbench) environment, CGA-PIDNN control system of threeinverted pendulum for three-dimensional simulation is designed based on hybridprogramming of MATLAB and virtual reality technology. It provides a new visual case forthe implementation of control system simulation.