Controlling a Double Inverted Pendulum Using Evolutionary Radial Basis Function Neural Network

A control method combining RBF neural networks with genetic algorithms is proposed, which controls a double inverted pendulum system. Genetic algorithms are used to search for the parameters of RBFNN controller and minimize the performance criterion of nonlinear optimization control constrained by the condition of the linear controller design. It is shown from the simulation results that the scheme is superior to state feedback and fuzzy control, possesses larger stable area (0θ 1,θ 225 °) and better interference rejection.