identification and Control of an Inverted Pendulum Using Neural Networks

Abstrac t- Neural networks can be used to identifY and control nonlinear mechanical systems. The objective in this paper is io control an inverted pendulum using a neural network as system identifier and two neural networks as controller. First, it is investigated whether neural networks are able to approximate the nonlinear behaviour of a mechanical system. The neural nets are optimized by using a second order optimization method and a prediction model. The inttial weights in the nets are scaled with the inputs of the net. Second, inverse neural models are used as a basis for the design of a classical computed torque controller. The controller is simulated on the inverted pendulum