Neuro-control of an inverted pendulum using Genetic Algorithm

The inverted pendulum is a highly nonlinear and open-loop unstable system. This means that standard linear techniques cannot exactly model and control the nonlinear dynamics of the system. This paper presents the neuro-control of an inverted pendulum using genetic algorithm. The system will be controlled via merging both neural networks and genetic algorithm. This paper focuses on training the neural network, through genetic algorithm, to identify a non-linear controller based on the nonlinear back-stepping control technique. In this paper, Artificial Neural Networks is trained by adaptive learning. The standard feed-forward ANN structure is used to model the controller of the inverted pendulum. This paper presents the simulation results of the controller found after Genetic Algorithm conversion; the system performance will then be compared to a nonlinear controller results. Finally, some important parameters in the Neural Network and Genetic Algorithm are changed to compare and assess their effect on the nonlinear controller (i.e. GA population size, number of generations, processing time, ...).