Control of an inverted pendulum using a neuro-fuzzy controller

In this paper, a form of neuro-fuzzy controller for the control of an inverted pendulum is being proposed. The membership functions of the fuzzy controller are implemented using multilayer pre-trained neural networks, thus enabling user to define arbitrarily the shape of the membership function by using small number of representative data set. The modularity of the controller enables off-line learning of the membership function using back-propagation multilayer feedforward network. The controller is implemented on the inverted pendulum hardware and found to work well. Self-organising features based on observing the trend of system kinetic and potential energy has also been proposed